**Ex-Post Clustering of Brazilian Beef Cattle Farms Using Soms and Cross-Evaluation Dea Models**

João Carlos Correia Baptista Soares de Mello, Eliane Gonçalves Gomes, Lidia Angulo Meza, Luiz Biondi Neto, Urbano Gomes Pinto de Abreu, Thiago Bernardino de Carvalho and Sergio de Zen

Additional information is available at the end of the chapter

http://dx.doi.org/10.5772/51324

**1. Introduction**

[16] Blanke, M., Kinnaert, M., Lunze, J., & Staroswicki, . (2003). *Diagnosis and Fault Toler‐*

[17] Benítez-Pérez, Héctor, & Benitez-Perez, Alma. (2010). The use of WAVELET strategy and Self Organizing Maps for Feature Extraction and Classification for Fault Diagno‐ sis. *International Journal of Innovative Computing, Information and Control, IJICIC*, 6(11),

*ant Control*, Springer-Verlag.

66 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

4923-4936.

The beef cattle production system is the set of technologies and management practices, ani‐ mal type, purpose of breeding, breed group and the eco-region where the activity is devel‐ oped. The central structure in the beef cattle production chain is the biological system of beef production, including the stages of creation (cow-calf production, stocker production, feedlot beef production) and their combinations. The cow-calf phase is the less profitable ac‐ tivity and the one that has the higher risk. However, it supports the entire structure of the production system.

Although it is undeniable that a systemic view in agriculture is important, it is not yet estab‐ lished in the Brazilian agricultural research. In this study, by using a non parametric techni‐ que known as Data Envelopment Analysis (DEA) and Self-Organizing Maps (SOMS), we intend to cluster cattle breeders of some Brazilian municipalities. The objective is to group the farmers according to their efficiency profiles regarding the decisions related to the com‐ position of their production systems, which are focused on the cow-calf phase. The farmers' decisions have direct impacts on the expenditures and on the income reached.

Efficiency is a relative concept: we compare the amount produced by a productive unit or firm given the available resources, to the maximum quantities that would have been pro‐ duced with the same amount of resources. There are different approaches to measure effi‐ ciency. The so-called parametric methods assume a pre-defined functional relationship between resources (inputs) and products (outputs). These methods usually use averages to

determine the amount that would have been produced. The non-parametric approaches, among which DEA, do not make any assumption about the functional form of the relation‐ ship between inputs and outputs. DEA assumes that the maximum that could have been produced is obtained in the sample under evaluation, by observing the firms that better per‐ form. DEA was initially presented in [1] and computes the efficiency of productive units, the so-called decision making units (DMUs). The DMUs, here the farms, use the same types of inputs to produce the same sort of outputs, and this set must be homogenous.

Our discussion proceeds as follows. In Section 2 we discuss the theoretical aspects of DEA and of cross-evaluation analysis. Section 3 is about the fundamental aspects of SOMS: Koho‐ nen Neural Networks. In Section 4 we present the proposed approach: DEA and SOMS in an ex-post evaluation. The case study is detailed in section 4. In Section 5 we discuss the re‐

Ex-Post Clustering of Brazilian Beef Cattle Farms Using Soms and Cross-Evaluation Dea Models

http://dx.doi.org/10.5772/51324

69

DEA is a mathematical programming approach developed to compute the comparative effi‐ ciency of productive units (firms or DMUs). The DMUs perform similar tasks and use differ‐ ent amounts of inputs to produce different quantities of outputs. In order to maximize the efficiency of each DMU, DEA models allow each one to choose the weight assigned to each variable in complete freedom, subject to some restrictions. In the case the firms under evalu‐ ation operate under different scales, it is possible to consider these differences in a DEA for‐ mulation. This model is known in the DEA literature as VRS (variable returns to scale) or BCC model (from Banker, Charnes and Cooper), as firstly discussed in [11]. On the other hand, if it is possible to assume that the DMUs operate under constant return to scale, i.e., there is a proportional relationship in increments or decrements between inputs and out‐ puts, the suitable model is the CRS (constant returns to scale) or CCR (from Charnes, Coop‐

The mathematical linear formulation for DEA CCR model is shown in model (1), where *h*1 is the efficiency of the DMU '1' under evaluation; *xik*and *y jk* are the *i-th* input and *j-*

are the output and the input weights or mul‐

(1)

sults. Finally, we summarize our conclusions and list the references.

**2. Dea and cross-efficiency evaluation**

er and Rhodes), as presented in the seminal paper [1].

max *<sup>h</sup>*<sup>1</sup> <sup>=</sup>∑

subject to

*vixi*<sup>1</sup> =1

*μj <sup>y</sup> jk* −∑ *i*=1 *r*

*vixik* ≤0

, *vi* ≥0, *k* =1...*n*, *j* =1...*s*, *i* =1...*r*

As previously mentioned, DEA allows each DMU to perform its self-evaluation, in other words, to choose its own set of multipliers in such a way that its efficiency score is maxi‐ mized. In [10], it was proposed that the optimal weights for each DMU can be used to evalu‐ ate its peers, i.e., to compute alternative efficiency scores for every other DMU. This evaluation performed by the complete set of DMUs is called cross-evaluation [12, 13] and the resulting measures of performance are the cross-efficiencies. In (2), *hk* <sup>1</sup>is the cross-effi‐

∑ *i*=1 *r*

∑ *j*=1 *s*

*μj*

*j*=1 *s μj y <sup>j</sup>*<sup>1</sup>

*th* output of the *k-th* DMU, *k*=1…*n*; *μj* and *vi*

tipliers, respectively.

Mathematically and in the presence of multiple inputs and multiple outputs, efficiency is the ratio between the weighted sum of outputs and the weighted sum of inputs. In the DEA ap‐ proach, each firm chooses the most appropriate set of weights with the view to maximize that ratio. This choice is neither arbitrary nor subjective, as it is based on some restrictions. The result of the ratio must mean efficiency, i.e., it must be a number between 0 and 1. Thus, it is necessary that the weights that a firm chooses when applied to itself and to the others (in total of *k* firms) cannot give a number bigger than one. These considerations are set in a mathematical programming problem that formalizes the DEA model. In this context, we may say that DEA scores are a benevolent measure of efficiency.

SOMS are a special case of neural networks. In the literature we can find some papers that discuss the joint use of neural networks and data envelopment analysis. SOMS are often used with DEA to perform an ex-ante clustering, in order to allow DEA to evaluate only ho‐ mogeneous units. For instance, in [2] Neural Networks and DEA were used to determine if the differences among efficiency scores were due to environmental variables or to the man‐ agement process. The use of Neural Network for clustering and benchmarking container ter‐ minals was done in [3]. In [4], authors used Kohonen self-organizing maps to cluster participating countries in the Olympics and then applied DEA for producing a new ranking of participating teams. In references [5, 6] they used the back propagation neural network algorithm to accelerate computations in DEA. In [7] authors applied Neural Networks to es‐ timate missing information for suppliers evaluation using DEA. In [8], SOMS were used to cluster the CEDERJ distance learning centers before applying DEA models to assess the effi‐ ciency of these centers. The differences between the clusters were taken into account by a homogenization of the centers for the purpose of comparing them in one single cluster.

In this chapter we propose a different approach. SOMS are used for ex-post evaluation, as discussed in [9]. In our approach we first perform a DEA evaluation and then we cluster the DMUs into groups based on DEA efficiencies. In our case we compute the farmers' efficien‐ cy profiles and the SOMS are used to group the DMUs according to their common character‐ istics. The efficiency profiles are derived from the DEA cross-efficiency matrix. In the crossevaluation approach [10] each DMU is assessed by its own set of multipliers as well as by the other DMUs' multipliers. Thus, the column vector of efficiencies calculated for each DMU in the cross-evaluation matrix is assumed as the efficiency profile. It is important to note that the DEA cross-efficiency technique is usually used for ranking purposes. Here, however, it is not used as a ranking tool but as the input to a SOM that will be used to clus‐ ter the units according to the way their efficiencies were obtained.

Our discussion proceeds as follows. In Section 2 we discuss the theoretical aspects of DEA and of cross-evaluation analysis. Section 3 is about the fundamental aspects of SOMS: Koho‐ nen Neural Networks. In Section 4 we present the proposed approach: DEA and SOMS in an ex-post evaluation. The case study is detailed in section 4. In Section 5 we discuss the re‐ sults. Finally, we summarize our conclusions and list the references.

## **2. Dea and cross-efficiency evaluation**

determine the amount that would have been produced. The non-parametric approaches, among which DEA, do not make any assumption about the functional form of the relation‐ ship between inputs and outputs. DEA assumes that the maximum that could have been produced is obtained in the sample under evaluation, by observing the firms that better per‐ form. DEA was initially presented in [1] and computes the efficiency of productive units, the so-called decision making units (DMUs). The DMUs, here the farms, use the same types of

Mathematically and in the presence of multiple inputs and multiple outputs, efficiency is the ratio between the weighted sum of outputs and the weighted sum of inputs. In the DEA ap‐ proach, each firm chooses the most appropriate set of weights with the view to maximize that ratio. This choice is neither arbitrary nor subjective, as it is based on some restrictions. The result of the ratio must mean efficiency, i.e., it must be a number between 0 and 1. Thus, it is necessary that the weights that a firm chooses when applied to itself and to the others (in total of *k* firms) cannot give a number bigger than one. These considerations are set in a mathematical programming problem that formalizes the DEA model. In this context, we

SOMS are a special case of neural networks. In the literature we can find some papers that discuss the joint use of neural networks and data envelopment analysis. SOMS are often used with DEA to perform an ex-ante clustering, in order to allow DEA to evaluate only ho‐ mogeneous units. For instance, in [2] Neural Networks and DEA were used to determine if the differences among efficiency scores were due to environmental variables or to the man‐ agement process. The use of Neural Network for clustering and benchmarking container ter‐ minals was done in [3]. In [4], authors used Kohonen self-organizing maps to cluster participating countries in the Olympics and then applied DEA for producing a new ranking of participating teams. In references [5, 6] they used the back propagation neural network algorithm to accelerate computations in DEA. In [7] authors applied Neural Networks to es‐ timate missing information for suppliers evaluation using DEA. In [8], SOMS were used to cluster the CEDERJ distance learning centers before applying DEA models to assess the effi‐ ciency of these centers. The differences between the clusters were taken into account by a homogenization of the centers for the purpose of comparing them in one single cluster.

In this chapter we propose a different approach. SOMS are used for ex-post evaluation, as discussed in [9]. In our approach we first perform a DEA evaluation and then we cluster the DMUs into groups based on DEA efficiencies. In our case we compute the farmers' efficien‐ cy profiles and the SOMS are used to group the DMUs according to their common character‐ istics. The efficiency profiles are derived from the DEA cross-efficiency matrix. In the crossevaluation approach [10] each DMU is assessed by its own set of multipliers as well as by the other DMUs' multipliers. Thus, the column vector of efficiencies calculated for each DMU in the cross-evaluation matrix is assumed as the efficiency profile. It is important to note that the DEA cross-efficiency technique is usually used for ranking purposes. Here, however, it is not used as a ranking tool but as the input to a SOM that will be used to clus‐

inputs to produce the same sort of outputs, and this set must be homogenous.

may say that DEA scores are a benevolent measure of efficiency.

68 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

ter the units according to the way their efficiencies were obtained.

DEA is a mathematical programming approach developed to compute the comparative effi‐ ciency of productive units (firms or DMUs). The DMUs perform similar tasks and use differ‐ ent amounts of inputs to produce different quantities of outputs. In order to maximize the efficiency of each DMU, DEA models allow each one to choose the weight assigned to each variable in complete freedom, subject to some restrictions. In the case the firms under evalu‐ ation operate under different scales, it is possible to consider these differences in a DEA for‐ mulation. This model is known in the DEA literature as VRS (variable returns to scale) or BCC model (from Banker, Charnes and Cooper), as firstly discussed in [11]. On the other hand, if it is possible to assume that the DMUs operate under constant return to scale, i.e., there is a proportional relationship in increments or decrements between inputs and out‐ puts, the suitable model is the CRS (constant returns to scale) or CCR (from Charnes, Coop‐ er and Rhodes), as presented in the seminal paper [1].

The mathematical linear formulation for DEA CCR model is shown in model (1), where *h*1 is the efficiency of the DMU '1' under evaluation; *xik*and *y jk* are the *i-th* input and *jth* output of the *k-th* DMU, *k*=1…*n*; *μj* and *vi* are the output and the input weights or mul‐ tipliers, respectively.

$$\begin{aligned} \max & \boldsymbol{h}\_{1} = \sum\_{j=1}^{s} \boldsymbol{\mu}\_{j} \mathbf{y}\_{j1} \\ \text{subject to} \\ & \sum\_{i=1}^{r} \boldsymbol{v}\_{i} \mathbf{x}\_{i1} = \mathbf{1} \\ & \sum\_{j=1}^{s} \boldsymbol{\mu}\_{j} \mathbf{y}\_{jk} - \sum\_{i=1}^{r} \boldsymbol{v}\_{i} \mathbf{x}\_{ik} \le \mathbf{0} \\ & \boldsymbol{\mu}\_{j} \quad \boldsymbol{v}\_{i} \ge \mathbf{0}, \quad k = 1 \dots n, \quad j = 1 \dots s, \quad i = 1 \dots r \end{aligned} \tag{1}$$

As previously mentioned, DEA allows each DMU to perform its self-evaluation, in other words, to choose its own set of multipliers in such a way that its efficiency score is maxi‐ mized. In [10], it was proposed that the optimal weights for each DMU can be used to evalu‐ ate its peers, i.e., to compute alternative efficiency scores for every other DMU. This evaluation performed by the complete set of DMUs is called cross-evaluation [12, 13] and the resulting measures of performance are the cross-efficiencies. In (2), *hk* <sup>1</sup>is the cross-effi‐ ciency of DMU '1' using the weights of the DMU '*k*'; *μ jk* is the weight of output *j* obtained for DMU '*k*'; *vik* is the weight of input *i* obtained for DMU '*k*'. The final cross-efficiency index is the average of all peer and self-evaluations.

$$\ln\_{\vec{k}\cdot\mathbf{1}} = \frac{\sum\_{j=1}^{s} \mu\_{jk} \mathbf{y}\_{j\mathbf{1}}}{\sum\_{i=1}^{r} \upsilon\_{ik} \mathbf{x}\_{i\mathbf{1}}} \tag{2}$$

**DMU 1 2 3 ...** *n* 1 *h11 h12 h13 ... h1n* 2 *h21 h22 h23 ... h2n* 3 *h31 h32 h33 ... h35* ... *... ... ... ... ... n hn1 hn2 hn3 ... hn5*

Ex-Post Clustering of Brazilian Beef Cattle Farms Using Soms and Cross-Evaluation Dea Models

http://dx.doi.org/10.5772/51324

71

We may point out that DMUs sharing similar weights distribution (with similar characteris‐ tics) will evaluate each other with high efficiency scores. On the contrary, those who have dissimilar characteristics will evaluate each other with low efficiency scores. The efficiency scores of a DMU (the column of the matrix) are limited by the efficiency score obtained in the self-evaluation (the diagonal of the matrix, *h kk*). In the self-evaluation, provided by the classic DEA models, a DMU is shown in the best possible way as to maximize its efficiency.

The human brain organizes information in a logic way. A paramount aspect of the self-or‐ ganized networks is motivated by the organization of the human brain in such a way that the sensory inputs are represented by topologically organized maps. The Kohonen self-or‐ ganizing map emulates the unsupervised learning in a simple and elegant way taking into

From topographic map development principle came up two feature mapping models: the model presented in [21, 22], having strong neurobiological motivations, and the Kohonen model [23], not as close to neurobiology as the previous one but enabling a simple comput‐ ing treatment stressing the essential characteristics of the brain maps. Moreover, the Koho‐

The SOMS are artificial neural networks special structures in a grid form that work in a sim‐ ilar way of the human brain, as far as the information organization is concerned, and are based on competitive learning. The most used SOM is the topologically interconnected twodimensional, where the neurons are represented by rectangular, hexagonal and random grid

In order to analyze the competitive process, let us suppose that the input space is *m*-dimen‐ sional and that *X* represents a random input pattern [24] such that one can write *X* = *X*1, *X*2, *X*3, ..., *Xm* . Assuming that the weight vector *W* of each neuron has the same dimension of the input space, for a given neuron *j* of a total of *l* neurons the weight vector

Therefore, all peer-evaluations will be lower or equal to that efficiency score.

**3. Fundamental aspects of som: kohonen neural network**

nen model or Kohonen Self-Organizing Map yields a low input dimension.

knots of neighbor neurons. Higher dimensional maps can also be modeled.

can be written as*W* = *W <sup>j</sup>*1, *W <sup>j</sup>*2, *W <sup>j</sup>*3, ..., *W jm* , *j*=1…*l*.

**Table 1.** Cross-efficiencies matrix for *n* DMUs.

account the neuron neighborhood [20].

As discussed in [14], the use of cross-evaluation has spread to a number of different areas, as in multiple criteria decision making to improve discrimination among alternatives, in a pref‐ erential election, in the ranking and selection of projects and technologies, among others. As discussed in [15-17] cross-evaluation commonly uses the CCR model due to the existence of negative efficiencies when choosing the BCC model. The usage of the BCC model for crossevaluation imposes the inclusion of an additional set of restrictions to ensure positive effi‐ ciencies [16, 17].

As was noticed in [10], in the DEA context the optimal set of weights for a DMU is not unique. Therefore it is necessary to choose one set of weights among all the possibilities. In the original cross-evaluation formulation these approaches led to nonlinear problems. In [12] two linear models were introduced, that are approximations to the original formulation. There are some other alternative models for the cross-efficiency evaluation. We mention the DEA-Game [18], among others [19]. These models may use two approaches. The first one, called "aggressive model", minimizes the cross-evaluation indexes of all DMUs. The second approach, known as "benevolent model", maximizes the cross-evaluation scores of all DMUs. The so-called aggressive *C <sup>k</sup>* formulation is shown in model (3), as presented in [12].

$$\begin{aligned} \min \, \mathcal{C}\_k &= \sum\_{j=1}^s \mu\_{jk} \sum\_{m \neq k} \mathcal{Y}\_{jm} \\ \text{subject to} \\ \sum\_{i=1}^r \upsilon\_{ik} \sum\_{m \neq k} \mathcal{X}\_{im} &= \mathbf{1} \\ \sum\_{j=1}^s \mu\_{jk} \mathcal{Y}\_{jm} - \sum\_{i=1}^r \upsilon\_{ik} \mathbf{x}\_{im} &\le \mathbf{0}, \quad \forall \ m \neq k \\ \sum\_{j=1}^s \mu\_{jk} \mathcal{Y}\_{jk} - \hbar\_{kk} \sum\_{i=1}^r \upsilon\_{ik} \mathbf{x}\_{ik} &= \mathbf{0} \\ \mu\_{jk} \mathcal{Y}\_{\cdot \cdot} \upsilon\_{jk} &\ge \mathbf{0} \end{aligned} \tag{3}$$

After determining the set of weights to be used in the cross-evaluation, a cross-efficiency matrix is calculated, as shown in Table 1. This matrix contains the self-evaluation (*h kk*) and the peer-evaluation (*h kl , k ≠ l*) of each DMU. The efficiency scores of a given DMU will be in the column of this matrix.


**Table 1.** Cross-efficiencies matrix for *n* DMUs.

ciency of DMU '1' using the weights of the DMU '*k*'; *μ jk* is the weight of output *j* obtained for DMU '*k*'; *vik* is the weight of input *i* obtained for DMU '*k*'. The final cross-efficiency index

(2)

(3)

*hk* <sup>1</sup> =

min *Ck* <sup>=</sup>∑

subject to

∑ *i*=1 *r vik*∑ *m*≠*k*

∑ *j*=1 *s*

∑ *j*=1 *s*

the column of this matrix.

*j*=1 *s*

*xim* =1

*i*=1 *r*

> *i*=1 *r*

*vik xim* ≤0, ∀*m*≠*k*

*vik xik* =0

After determining the set of weights to be used in the cross-evaluation, a cross-efficiency matrix is calculated, as shown in Table 1. This matrix contains the self-evaluation (*h kk*) and the peer-evaluation (*h kl , k ≠ l*) of each DMU. The efficiency scores of a given DMU will be in

*<sup>μ</sup> jk <sup>y</sup> jm* −∑

*<sup>μ</sup> jk <sup>y</sup> jk* <sup>−</sup>*hkk*∑

*μ jk* , *vik* ≥0

*<sup>μ</sup> jk*∑ *m*≠*k y jm*

∑ *<sup>j</sup>*=1 *<sup>s</sup> <sup>μ</sup> jk <sup>y</sup> <sup>j</sup>*<sup>1</sup>

∑*<sup>i</sup>*=1 *<sup>r</sup> vik xi*<sup>1</sup>

As discussed in [14], the use of cross-evaluation has spread to a number of different areas, as in multiple criteria decision making to improve discrimination among alternatives, in a pref‐ erential election, in the ranking and selection of projects and technologies, among others. As discussed in [15-17] cross-evaluation commonly uses the CCR model due to the existence of negative efficiencies when choosing the BCC model. The usage of the BCC model for crossevaluation imposes the inclusion of an additional set of restrictions to ensure positive effi‐

As was noticed in [10], in the DEA context the optimal set of weights for a DMU is not unique. Therefore it is necessary to choose one set of weights among all the possibilities. In the original cross-evaluation formulation these approaches led to nonlinear problems. In [12] two linear models were introduced, that are approximations to the original formulation. There are some other alternative models for the cross-efficiency evaluation. We mention the DEA-Game [18], among others [19]. These models may use two approaches. The first one, called "aggressive model", minimizes the cross-evaluation indexes of all DMUs. The second approach, known as "benevolent model", maximizes the cross-evaluation scores of all DMUs. The so-called aggressive *C <sup>k</sup>* formulation is shown in model (3), as presented in [12].

is the average of all peer and self-evaluations.

70 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

ciencies [16, 17].

We may point out that DMUs sharing similar weights distribution (with similar characteris‐ tics) will evaluate each other with high efficiency scores. On the contrary, those who have dissimilar characteristics will evaluate each other with low efficiency scores. The efficiency scores of a DMU (the column of the matrix) are limited by the efficiency score obtained in the self-evaluation (the diagonal of the matrix, *h kk*). In the self-evaluation, provided by the classic DEA models, a DMU is shown in the best possible way as to maximize its efficiency. Therefore, all peer-evaluations will be lower or equal to that efficiency score.

## **3. Fundamental aspects of som: kohonen neural network**

The human brain organizes information in a logic way. A paramount aspect of the self-or‐ ganized networks is motivated by the organization of the human brain in such a way that the sensory inputs are represented by topologically organized maps. The Kohonen self-or‐ ganizing map emulates the unsupervised learning in a simple and elegant way taking into account the neuron neighborhood [20].

From topographic map development principle came up two feature mapping models: the model presented in [21, 22], having strong neurobiological motivations, and the Kohonen model [23], not as close to neurobiology as the previous one but enabling a simple comput‐ ing treatment stressing the essential characteristics of the brain maps. Moreover, the Koho‐ nen model or Kohonen Self-Organizing Map yields a low input dimension.

The SOMS are artificial neural networks special structures in a grid form that work in a sim‐ ilar way of the human brain, as far as the information organization is concerned, and are based on competitive learning. The most used SOM is the topologically interconnected twodimensional, where the neurons are represented by rectangular, hexagonal and random grid knots of neighbor neurons. Higher dimensional maps can also be modeled.

In order to analyze the competitive process, let us suppose that the input space is *m*-dimen‐ sional and that *X* represents a random input pattern [24] such that one can write *X* = *X*1, *X*2, *X*3, ..., *Xm* . Assuming that the weight vector *W* of each neuron has the same dimension of the input space, for a given neuron *j* of a total of *l* neurons the weight vector can be written as*W* = *W <sup>j</sup>*1, *W <sup>j</sup>*2, *W <sup>j</sup>*3, ..., *W jm* , *j*=1…*l*.

For each input vector, the scalar product is evaluated in order to find the *X* vector which is closest to the weight vector *W*. By comparison, the maximum scalar product max(*Wj <sup>t</sup>* ⋅ *X* ) is chosen, representing the location in which the topological neighborhood of excited neurons should be centered. Maximizing this scalar product is equivalent to minimizing the Euclidi‐ an distance between *X* and *W*. Other metrics could also be used.

The neuron with the weight vector closest to the input vector *X* is called the winner neuron, whose index *V* (*X* ) is given by*V* (*X* )=min *X* −*Wj* . In the cooperative process the winner neuron locates the centre of a topological neighborhood of cooperating neurons. The active winner tends to strongly stimulate its closest neighbor neurons and weakly the farthest ones. It is essential to find a topological neighborhood function *N <sup>j</sup>*,*<sup>V</sup>* (*<sup>X</sup>* ) that is independent from the winner neuron location. That neighborhood function should represent the topolog‐ ical neighborhood centered in the winner neuron, denoted by *V*, having as closest lateral neighbors a group of excited and cooperative neurons, from which a representative *j* neuron can be chosen. The lateral distance *Dj*,*V* between the winner neuron, indexed by *V*, and the excited neuron, indexed by *j*, can be written as in (4), where is the neighborhood width [24].

$$N\_{\boldsymbol{j},V(X)} = \exp\left(-\frac{D\_{\boldsymbol{j},V}^2}{2\sigma^2}\right) \tag{4}$$

The adaptive process is the last phase of the SOM procedure. During this phase is carried out the adjustment of the connection weights of the neurons. In order the network succeed in the self-organization task, it is necessary to update the weights *Wj* of the excited j neuron

The change to the weight vector of the excited neuron j in the grid can be written as

*ΔWj* =*η N <sup>j</sup>*,*<sup>V</sup>* (*<sup>X</sup>* )(*X* −*Wj*

(*n*) + *η*(*n*) *N <sup>j</sup>*,*<sup>V</sup>* (*<sup>X</sup>* )

Using discrete-time notation, a weight updating equation can be written, which applies to all neurons that are within the topographic neighborhood equation of the winner neuron

In (9) the learning rate parameter changes in each iteration, with an initial value around 0.1 and decreases with increasing discrete-time t up to values above 0.01 [20]. To that end, equa‐ tion (10) is written in which decays exponentially and *τ*2 is another time-constant of the SOM algorithm. For the fulfillment of the requirements one could choose for instance,

As discussed in Section 2, in the cross-evaluation approach each DMU is assessed by its own set of multipliers (self-evaluation) as well as by the other DMUs' multipliers (peer-evalua‐ tion). The cross-efficiency matrix provides the cross-efficiencies of every DMU, each column providing the efficiencies of each DMU, i.e., is the evaluation performed by the other DMUs. As previously discussed, DMUs with the same characteristics will evaluate each other with high efficiencies, or with low efficiencies otherwise. Therefore, we can assume that each col‐

When high value of weights are assigned to the variables of DMU A this implies that DMU A will have a good efficiency score (ratio between the weighted sum of the outputs and the weighted sum of the inputs). In the context of cross-evaluation, when these weights are as‐

, where is the learning rate parameter, *η yjX* is the Hebbian term, and

Ex-Post Clustering of Brazilian Beef Cattle Farms Using Soms and Cross-Evaluation Dea Models

(*n*)(*X* −*Wj*

) is chosen as*g*(*yj*

http://dx.doi.org/10.5772/51324

) (8)

(*t*)) (9)

), *t* =0, 1, 2, 3, ... (10)

) =*η yj* . If 73

relatively to the input vector X.

)*Wj*

is called forgetting term [23-25].

*Wj*

(*n* + 1)=*Wj*

*<sup>η</sup>*(*t*)=*η*0exp( <sup>−</sup> *<sup>t</sup>*

umn of the cross-evaluation matrix is the efficiency profile of each DMU.

**4. Dea and soms in an ex-post evaluation**

*τ*2

*yj* = *N <sup>j</sup>*,*<sup>V</sup>* (*<sup>X</sup>* ), then the expression can be written as (8).

In order to satisfy the requirement, a linear function for *g*(*yj*

*ΔWj* =*η yjX* − *g*(*yj*

[23, 24], as in (9).

*η*<sup>0</sup> =0.1and*τ*<sup>2</sup> =1, 000.

*g*(*yj* )*Wj*

The more dependent is the lateral distance*Dj*,*<sup>V</sup>* , the greater will be the cooperation among the neighborhood neurons. So, for a two-dimensional output grid, the lateral distance can be defined as in (5), for which the discrete vector ℘ *<sup>j</sup>* represents the position of the excited neu‐ ron, and ℘*V* the position of the neuron that won the competition.

$$\|D\\_\_{j,V} = \sqrt{\|\:\wp\_j - \wp\_V\|\:\vert^2} \tag{5}$$

The topological neighborhood should decrease with discrete time t. In order to accomplish that, the width *σ* of the topological neighborhood *N <sup>j</sup>*,*<sup>V</sup>* (*<sup>X</sup>* ) should decrease in time. The width could be written as in (6), where *σ*0 represents the initial value of the neighborhood width and *τ*<sup>1</sup> a time constant. Usually *σ*<sup>0</sup> is adjusted to have the same value as the grid ratio, i.e.,*τ*<sup>1</sup> =1, 000 / log*σ*0.

$$\sigma\_0(t) = \sigma\_0 \exp\left(-\frac{t}{\tau\_1}\right) \quad t = 0, \ 1, \ 2, \ 3, \ \dots \tag{6}$$

The expression of the topological neighborhood in time can be written as in (7).

$$N\_{\boldsymbol{f},V(X)}(t) = \exp\left(-\frac{D\_{\boldsymbol{f},V}^2}{2\sigma^2(t)}\right), \quad t = 0, \ 1, \ 2, \ 3, \ \dots \tag{7}$$

The adaptive process is the last phase of the SOM procedure. During this phase is carried out the adjustment of the connection weights of the neurons. In order the network succeed in the self-organization task, it is necessary to update the weights *Wj* of the excited j neuron relatively to the input vector X.

The change to the weight vector of the excited neuron j in the grid can be written as *ΔWj* =*η yjX* − *g*(*yj* )*Wj* , where is the learning rate parameter, *η yjX* is the Hebbian term, and *g*(*yj* )*Wj* is called forgetting term [23-25].

In order to satisfy the requirement, a linear function for *g*(*yj* ) is chosen as*g*(*yj* ) =*η yj* . If *yj* = *N <sup>j</sup>*,*<sup>V</sup>* (*<sup>X</sup>* ), then the expression can be written as (8).

$$
\Delta \mathcal{W}\_{\hat{j}} = \eta \text{ } \mathcal{N}\_{\hat{j}, V(X)} \text{( $X - \mathcal{W}\_{\hat{j}}$ )}\tag{8}
$$

Using discrete-time notation, a weight updating equation can be written, which applies to all neurons that are within the topographic neighborhood equation of the winner neuron [23, 24], as in (9).

$$\mathcal{W}\_j(n+1) = \mathcal{W}\_j(n) + \eta(n) \text{ N}\_{j,V(X)}(n) \text{( $X - \mathcal{W}\_j(t)$ )}\tag{9}$$

In (9) the learning rate parameter changes in each iteration, with an initial value around 0.1 and decreases with increasing discrete-time t up to values above 0.01 [20]. To that end, equa‐ tion (10) is written in which decays exponentially and *τ*2 is another time-constant of the SOM algorithm. For the fulfillment of the requirements one could choose for instance, *η*<sup>0</sup> =0.1and*τ*<sup>2</sup> =1, 000.

$$\eta(t) = \eta\_0 \exp\left(-\frac{t}{\tau\_2}\right), \quad t = 0, \ 1, \ 2, \ 3, \ \dots \tag{10}$$

#### **4. Dea and soms in an ex-post evaluation**

For each input vector, the scalar product is evaluated in order to find the *X* vector which is

chosen, representing the location in which the topological neighborhood of excited neurons should be centered. Maximizing this scalar product is equivalent to minimizing the Euclidi‐

The neuron with the weight vector closest to the input vector *X* is called the winner neuron, whose index *V* (*X* ) is given by*V* (*X* )=min *X* −*Wj* . In the cooperative process the winner neuron locates the centre of a topological neighborhood of cooperating neurons. The active winner tends to strongly stimulate its closest neighbor neurons and weakly the farthest ones. It is essential to find a topological neighborhood function *N <sup>j</sup>*,*<sup>V</sup>* (*<sup>X</sup>* ) that is independent from the winner neuron location. That neighborhood function should represent the topolog‐ ical neighborhood centered in the winner neuron, denoted by *V*, having as closest lateral neighbors a group of excited and cooperative neurons, from which a representative *j* neuron can be chosen. The lateral distance *Dj*,*V* between the winner neuron, indexed by *V*, and the excited neuron, indexed by *j*, can be written as in (4), where is the neighborhood width [24].

*<sup>t</sup>* ⋅ *X* ) is

closest to the weight vector *W*. By comparison, the maximum scalar product max(*Wj*

*N <sup>j</sup>*,*<sup>V</sup>* (*<sup>X</sup>* )=exp( −

*Dj*,*<sup>V</sup>* = ℘ *<sup>j</sup>* −℘*<sup>V</sup>*

*τ*1

The expression of the topological neighborhood in time can be written as in (7).

2 2*σ* 2(*t*)

(*t*)=exp( <sup>−</sup> *Dj*,*<sup>V</sup>*

defined as in (5), for which the discrete vector ℘ *<sup>j</sup>*

i.e.,*τ*<sup>1</sup> =1, 000 / log*σ*0.

ron, and ℘*V* the position of the neuron that won the competition.

*<sup>σ</sup>*(*t*)=*σ*0exp( <sup>−</sup> *<sup>t</sup>*

*N <sup>j</sup>*,*<sup>V</sup>* (*<sup>X</sup>* )

*Dj*,*<sup>V</sup>* 2

The more dependent is the lateral distance*Dj*,*<sup>V</sup>* , the greater will be the cooperation among the neighborhood neurons. So, for a two-dimensional output grid, the lateral distance can be

The topological neighborhood should decrease with discrete time t. In order to accomplish that, the width *σ* of the topological neighborhood *N <sup>j</sup>*,*<sup>V</sup>* (*<sup>X</sup>* ) should decrease in time. The width could be written as in (6), where *σ*0 represents the initial value of the neighborhood width and *τ*<sup>1</sup> a time constant. Usually *σ*<sup>0</sup> is adjusted to have the same value as the grid ratio,

<sup>2</sup>*<sup>σ</sup>* <sup>2</sup> ) (4)

represents the position of the excited neu‐

<sup>2</sup> (5)

), *t* =0, 1, 2, 3, ... (6)

), *t* =0, 1, 2, 3, ... (7)

an distance between *X* and *W*. Other metrics could also be used.

72 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

As discussed in Section 2, in the cross-evaluation approach each DMU is assessed by its own set of multipliers (self-evaluation) as well as by the other DMUs' multipliers (peer-evalua‐ tion). The cross-efficiency matrix provides the cross-efficiencies of every DMU, each column providing the efficiencies of each DMU, i.e., is the evaluation performed by the other DMUs. As previously discussed, DMUs with the same characteristics will evaluate each other with high efficiencies, or with low efficiencies otherwise. Therefore, we can assume that each col‐ umn of the cross-evaluation matrix is the efficiency profile of each DMU.

When high value of weights are assigned to the variables of DMU A this implies that DMU A will have a good efficiency score (ratio between the weighted sum of the outputs and the weighted sum of the inputs). In the context of cross-evaluation, when these weights are as‐ signed to DMU B and it also has high efficiency score we may say that DMU B has similar characteristics to DMU A. We can observe the case in which a DMU have a good perform‐ ance in all variables (or in the majority of them), but it is not unusual the case where the efficiency score is computed based only on two variables (one input and one output). This occurs when the DMU have a low performance on some variables and null weights (or near zero) are assigned to them. Therefore, considering DMUs A and B, DMU A may evaluate DMU B positively, but when DMU B evaluates DMU A, its efficiency may be low. In other words, cross-evaluation may not bring the same results for both.

ciencies and cluster analysis to group the nations and to select more appropriate targets for

Ex-Post Clustering of Brazilian Beef Cattle Farms Using Soms and Cross-Evaluation Dea Models

http://dx.doi.org/10.5772/51324

75

The beef complex in Brazil is consolidated as an important link in the production and in the international trade: Brazil is the largest exporter and the second largest producer of beef. Considering this scenario, the study and the evaluation of beef cattle production systems are

In Brazil, the cow-calf beef cattle phase occurs predominantly in extensive continuous graz‐ ing, with native or cultivated pastures, encompassing: calves (until weaning or even one year old), cows, heifers and bulls. The cow-calf phase supports the entire structure of the

This case study seeks to assess the comparative performance of extensive livestock modal production systems in its cow-calf phase, in some Brazilian municipalities. The objective is to measure their performance and to group them according to the decisions regarding the composition of the production system. This has a direct impact on the expenditures and on

Primary data were collected through the panel system, which allows the definition of repre‐

In this approach, based on the experience of the participating farmers, it is characterized the property that is the most commonly found in the region, i.e., a property or a production system that is representative of the locality under study. In some cases it is not possible to determine this typology and more than one representative property or production system are specified. The panel is a less costly procedure of obtaining information than the census or the sam‐ pling of farms. The technique is applied during a meeting with a group of one or more re‐ searchers, one technician and eight regional farmers. Meetings are scheduled in advance, with the support of rural unions and regional contacts. At the end of that debate one can say that any characterization of the typical farm in the region has the consent of the farmers. Thus, productivity rates, establishment costs, fixed and variable costs, i.e., all the numbers

It is noteworthy that the rates and the costs reported by each participant are not related to their properties, but with a single farm, declared at the beginning of the panel as the one that best represents the scale of operation and the production system of most of the local properties. This study evaluated 21 beef cattle modal production systems that performed only the cowcalf phase, in seven states of Brazil. The data, derived from the indicators of the project de‐

resulting from the panel, tend to be fairly close to the regional reality.

poorly performing countries to use as benchmarks.

important tools for enhancing the performance of this sector.

**5. Case study**

beef production chain.

the income generated.

**5.1 Data Source**

sentative farms, as proposed in [27].

In this chapter we will use SOMS to cluster the farmers with similar characteristics using the cross-efficiency matrix information. However, the cross-efficiencies are dependent on the DMU self-evaluation. The cross-efficiency scores may be smaller compared to those ob‐ tained by other DMUs (lines in the matrix), but they may be consistent or very similar to the efficiency score obtained by the DMU (column in the matrix). Therefore, as our clusters will be defined by the efficiency profile, we need to remove the self-evaluation effect of each DMU. In this approach this is accomplished by normalizing each column of the matrix by the self-evaluation, i.e., all values are divided by the CCR DEA efficiency of the DMU. This is shown in Table 2. The self-evaluation is the highest efficiency score of the DMU, located in the diagonal of the cross-evaluation matrix. As a consequence, we will have unitary values in the matrix diagonal. The objective of this operation is to group the DMUs by the way they are evaluated by the others rather than by their efficiency level. This minimizes the benevo‐ lent characteristic of the classic DEA models.


**Table 2.** Normalized cross-efficiency matrix.

In the traditional cross-evaluation applications, the resulting average cross-efficiency index is commonly used for ranking purposes. Here, in contrast, the square matrix of normalized efficiencies will be submitted to the SOMS procedure (as its input) to generate the homoge‐ neous clusters. In other words, the normalized cross-evaluation matrix will be used to clus‐ ter the DMUs according to their efficiency profile.

An approach based on cross-evaluation and cluster technique was used in [26] to group par‐ ticipating countries in Summer Olympics Games. The authors used the average cross-effi‐ ciencies and cluster analysis to group the nations and to select more appropriate targets for poorly performing countries to use as benchmarks.

## **5. Case study**

signed to DMU B and it also has high efficiency score we may say that DMU B has similar characteristics to DMU A. We can observe the case in which a DMU have a good perform‐ ance in all variables (or in the majority of them), but it is not unusual the case where the efficiency score is computed based only on two variables (one input and one output). This occurs when the DMU have a low performance on some variables and null weights (or near zero) are assigned to them. Therefore, considering DMUs A and B, DMU A may evaluate DMU B positively, but when DMU B evaluates DMU A, its efficiency may be low. In other

In this chapter we will use SOMS to cluster the farmers with similar characteristics using the cross-efficiency matrix information. However, the cross-efficiencies are dependent on the DMU self-evaluation. The cross-efficiency scores may be smaller compared to those ob‐ tained by other DMUs (lines in the matrix), but they may be consistent or very similar to the efficiency score obtained by the DMU (column in the matrix). Therefore, as our clusters will be defined by the efficiency profile, we need to remove the self-evaluation effect of each DMU. In this approach this is accomplished by normalizing each column of the matrix by the self-evaluation, i.e., all values are divided by the CCR DEA efficiency of the DMU. This is shown in Table 2. The self-evaluation is the highest efficiency score of the DMU, located in the diagonal of the cross-evaluation matrix. As a consequence, we will have unitary values in the matrix diagonal. The objective of this operation is to group the DMUs by the way they are evaluated by the others rather than by their efficiency level. This minimizes the benevo‐

**DMU 1 2 3 ...** *n*

... *... ... ... ... ...*

In the traditional cross-evaluation applications, the resulting average cross-efficiency index is commonly used for ranking purposes. Here, in contrast, the square matrix of normalized efficiencies will be submitted to the SOMS procedure (as its input) to generate the homoge‐ neous clusters. In other words, the normalized cross-evaluation matrix will be used to clus‐

An approach based on cross-evaluation and cluster technique was used in [26] to group par‐ ticipating countries in Summer Olympics Games. The authors used the average cross-effi‐

*h13 h33*

*h23 h33*

*h33 h33*

*hn13 h33*

*... h1n*

*... h2n*

*... h3n*

*... hnn*

*hnn*

*hnn*

*hnn*

*hnn*

*h12 h22*

*h22 h22*

*h32 h22*

*hn2 h22*

words, cross-evaluation may not bring the same results for both.

lent characteristic of the classic DEA models.

74 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

*h11 h11*

*h21 h11*

*h31 h11*

*<sup>n</sup> hn1 h11*

ter the DMUs according to their efficiency profile.

1

2

3

**Table 2.** Normalized cross-efficiency matrix.

The beef complex in Brazil is consolidated as an important link in the production and in the international trade: Brazil is the largest exporter and the second largest producer of beef. Considering this scenario, the study and the evaluation of beef cattle production systems are important tools for enhancing the performance of this sector.

In Brazil, the cow-calf beef cattle phase occurs predominantly in extensive continuous graz‐ ing, with native or cultivated pastures, encompassing: calves (until weaning or even one year old), cows, heifers and bulls. The cow-calf phase supports the entire structure of the beef production chain.

This case study seeks to assess the comparative performance of extensive livestock modal production systems in its cow-calf phase, in some Brazilian municipalities. The objective is to measure their performance and to group them according to the decisions regarding the composition of the production system. This has a direct impact on the expenditures and on the income generated.

## **5.1 Data Source**

Primary data were collected through the panel system, which allows the definition of repre‐ sentative farms, as proposed in [27].

In this approach, based on the experience of the participating farmers, it is characterized the property that is the most commonly found in the region, i.e., a property or a production system that is representative of the locality under study. In some cases it is not possible to determine this typology and more than one representative property or production system are specified.

The panel is a less costly procedure of obtaining information than the census or the sam‐ pling of farms. The technique is applied during a meeting with a group of one or more re‐ searchers, one technician and eight regional farmers. Meetings are scheduled in advance, with the support of rural unions and regional contacts. At the end of that debate one can say that any characterization of the typical farm in the region has the consent of the farmers. Thus, productivity rates, establishment costs, fixed and variable costs, i.e., all the numbers resulting from the panel, tend to be fairly close to the regional reality.

It is noteworthy that the rates and the costs reported by each participant are not related to their properties, but with a single farm, declared at the beginning of the panel as the one that best represents the scale of operation and the production system of most of the local properties.

This study evaluated 21 beef cattle modal production systems that performed only the cowcalf phase, in seven states of Brazil. The data, derived from the indicators of the project de‐ veloped by the Centro de Estudos Avançados em Economia Aplicada and the Confederação da Agricultura e Pecuária do Brasil, were collected in municipalities of these seven states: Mato Grosso do Sul - MS (eight), Goiás - GO (four), Rio Grande do Sul - RS (one), Minas Gerais - MG (four), Tocantins - TO (two), São Paulo - SP (one) and Bahia - BA (one). Panels with the farmers, with the support of the local rural technical assistance, were performed to collect the data, according to the methodology described in [28].

**5.2.2 Variables**

in ranches with herds of genetic selection.

cattle production system.

**6. Results and discussion**

rons against the 21 DMUs of the data set.

efficiency scores [17].

The technicians and the researchers mentioned in 5.1, analyzed the variables set and imme‐ diately identified those relevant to our study. They selected "number of bulls" as the input variable, since this variable represents a significant portion of all total expenditures of the ranchers that produce calves. It is directly linked to the quality of animals that will be sold in these systems. This is also the only category that is purchased from other herd, especially

Ex-Post Clustering of Brazilian Beef Cattle Farms Using Soms and Cross-Evaluation Dea Models

http://dx.doi.org/10.5772/51324

77

The products of the system that are responsible for the main revenue from the cow-calf sys‐ tems were chosen as the output variables. These are the "number of calves on the herd" and the "number of cull cows". All calves produced are sold on the property and generate in‐ come. Cull cows are those that are sold, as they are not part of the production system either by higher age or by reproductive performance lower than desired. The decision to fit be‐ tween these variables is important because it will provide the dynamics structure of the breeding herd. This is the fundamental factor for keeping the economic viability of the beef

The variables indicated by experts need to be examined by analysts to determine whether they conform to the properties required by the DEA models. In particular, there must be a causal relationship between each input-output pair [29]. There is a clear causal relationship between the output "number of calves on the herd" and the input "number of bulls". The same cannot be said of the relationship between the input and the output "number of cull cows". Actually, there is no direct causal relationship between these variables; however there is a cost-benefit relationship. In the case the rancher has a big number of bulls (that represent an expense) he must earn more, either through the sale of calves or cows. There‐ fore, the "bulls – cull cows" ratio makes sense when using DEA to analyze cost-benefit ra‐ tios, and not just pure productive relations. This interpretation of DEA was introduced in

Figure 1 shows the cross-evaluation matrix. It was computed using the SIAD software [34]. As previously mentioned, we used the DEA CCR as the BCC model can generate negative

The values of the Figure 1, cross-evaluation matrix, were normalized before using SOMS, as explained in Section 4. Four different grid dimensions were tested: the 5x5 grid, the 3x3 grid, the 2x2 grid and the 2x1 grid. We began our analysis with the 5x5 grid since it yields 25 neu‐

[30] and was used in [31, 32]. Generalizations of this usage can be seen in [33].

## **5.2 Modeling**

#### **5.2.1 DMUs**

The objective of the DEA model is to measure the performance of the farmers' decision re‐ garding the composition of the rearing production system. Thereby, the DMUs are the 21 modal systems, identified from the panel discussions in 21 cities in seven Brazilian states. Table 3 presents the dataset.


**Table 3.** DMUs, inputs and outputs.

### **5.2.2 Variables**

veloped by the Centro de Estudos Avançados em Economia Aplicada and the Confederação da Agricultura e Pecuária do Brasil, were collected in municipalities of these seven states: Mato Grosso do Sul - MS (eight), Goiás - GO (four), Rio Grande do Sul - RS (one), Minas Gerais - MG (four), Tocantins - TO (two), São Paulo - SP (one) and Bahia - BA (one). Panels with the farmers, with the support of the local rural technical assistance, were performed to

The objective of the DEA model is to measure the performance of the farmers' decision re‐ garding the composition of the rearing production system. Thereby, the DMUs are the 21 modal systems, identified from the panel discussions in 21 cities in seven Brazilian states.

> **Municipality State Code (input) (output) (output)** Alvorada TO #1 12 147 30 Amambai MS #2 15 143 40 Aquidauana MS #3 92 713 214 Bonito MS #4 14 166 75 Brasilândia MS #5 31 290 178 Camapuã MS #6 9 65 33 Carlos Chagas MG #7 19 297 160 Catalão GO #8 8 81 42 Corumbá MS #9 69 455 200 Itamarajú BA #10 4 44 18 Lavras do Sul RS #11 5 58 30 Montes Claros MG #12 5 47 28 Niquelândia GO #13 4 35 18 Paraíso do Tocantins TO #14 12 123 35 Porangatu GO #15 5 46 23 Ribas Rio Pardo MS #16 15 143 70 Rio Verde GO #17 23 196 82 São Gabriel d'Oeste MS #18 11 95 40 Tupã SP #19 5 46 30 Uberaba MG #20 5 66 36 Uberlândia MG #21 2 20 10

**DMUs Breeders Calves Cull cows**

collect the data, according to the methodology described in [28].

76 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

**5.2 Modeling**

Table 3 presents the dataset.

**Table 3.** DMUs, inputs and outputs.

**5.2.1 DMUs**

The technicians and the researchers mentioned in 5.1, analyzed the variables set and imme‐ diately identified those relevant to our study. They selected "number of bulls" as the input variable, since this variable represents a significant portion of all total expenditures of the ranchers that produce calves. It is directly linked to the quality of animals that will be sold in these systems. This is also the only category that is purchased from other herd, especially in ranches with herds of genetic selection.

The products of the system that are responsible for the main revenue from the cow-calf sys‐ tems were chosen as the output variables. These are the "number of calves on the herd" and the "number of cull cows". All calves produced are sold on the property and generate in‐ come. Cull cows are those that are sold, as they are not part of the production system either by higher age or by reproductive performance lower than desired. The decision to fit be‐ tween these variables is important because it will provide the dynamics structure of the breeding herd. This is the fundamental factor for keeping the economic viability of the beef cattle production system.

The variables indicated by experts need to be examined by analysts to determine whether they conform to the properties required by the DEA models. In particular, there must be a causal relationship between each input-output pair [29]. There is a clear causal relationship between the output "number of calves on the herd" and the input "number of bulls". The same cannot be said of the relationship between the input and the output "number of cull cows". Actually, there is no direct causal relationship between these variables; however there is a cost-benefit relationship. In the case the rancher has a big number of bulls (that represent an expense) he must earn more, either through the sale of calves or cows. There‐ fore, the "bulls – cull cows" ratio makes sense when using DEA to analyze cost-benefit ra‐ tios, and not just pure productive relations. This interpretation of DEA was introduced in [30] and was used in [31, 32]. Generalizations of this usage can be seen in [33].

## **6. Results and discussion**

Figure 1 shows the cross-evaluation matrix. It was computed using the SIAD software [34]. As previously mentioned, we used the DEA CCR as the BCC model can generate negative efficiency scores [17].

The values of the Figure 1, cross-evaluation matrix, were normalized before using SOMS, as explained in Section 4. Four different grid dimensions were tested: the 5x5 grid, the 3x3 grid, the 2x2 grid and the 2x1 grid. We began our analysis with the 5x5 grid since it yields 25 neu‐ rons against the 21 DMUs of the data set.


The 5x5 grid provided the results shown in Table 4. There were obtained 18 clusters and al‐ most all contained one municipality (DMU). The exceptions were cluster 5 (which contains Brasilândia and Montes Claros), cluster 14 (compounded of Porangatu and Ribas Rio Par‐

Ex-Post Clustering of Brazilian Beef Cattle Farms Using Soms and Cross-Evaluation Dea Models

http://dx.doi.org/10.5772/51324

79

Figure 2 shows the resulting allocation for the 5x5 grid. The clusters with two DMUs were the ones in which the municipalities had the most similar efficiency profiles, i.e., strongest similar‐ ities. If these six DMUs hadn't met their similar pairs they would had been allocated in the empty clusters, each DMU in one cluster, instead of being grouped in pairs in three clusters. These three clusters were formed mainly due to the similarity of the decision profile in re‐ gards to the cows' culling age in the production systems of Brasilândia and Montes Claros (8.2 and 8.5 years), Ribas Rio Pardo and Porangatu (9.2 and 9.0 years), Rio Verde and São Gabriel d'Oeste (10.3 and 10.7 years). In practice, however, the 5x5 grid didn't perform well, as it hadn't grouped other production systems with similar profiles. Thus, this grid didn't provide the basis for an accurate synthesis of convergence points among the other systems

The 5x5 grid allocated many single municipalities in the clusters, i.e., one DMU in one clus‐ ter. Therefore, this topology was not suitable for this study. We also analyzed the 4x4 grid. We inferred that the 4x4 typology was also not adequate for this study as we obtained simi‐

The set shown in Table 5 is from the 3x3 grid. We observe that some DMUs were in the same clusters they were allocated in the 5x5 grid. This was the case of the cluster composed of Brasilândia and Montes Claros. In this cluster were added the DMUs Catalão, Ribas Rio Pardo, Tupã and Uberlândia. The same occurred with the group Rio Verde – São Gabriel

do), and cluster 15 (with Rio Verde and São Gabriel d'Oeste).

under assessment.

lar results to the 5x5 grid.

**Figure 2.** DMUs' distribution for the 5x5 grid.

**Figure 1.** Cross-evaluation matrix. In the diagonal are the scores computed by the self-evaluations. Other cells show the peer-evaluations.


**Table 4.** Clusters obtained with the 5x5 grid.

The 5x5 grid provided the results shown in Table 4. There were obtained 18 clusters and al‐ most all contained one municipality (DMU). The exceptions were cluster 5 (which contains Brasilândia and Montes Claros), cluster 14 (compounded of Porangatu and Ribas Rio Par‐ do), and cluster 15 (with Rio Verde and São Gabriel d'Oeste).

Figure 2 shows the resulting allocation for the 5x5 grid. The clusters with two DMUs were the ones in which the municipalities had the most similar efficiency profiles, i.e., strongest similar‐ ities. If these six DMUs hadn't met their similar pairs they would had been allocated in the empty clusters, each DMU in one cluster, instead of being grouped in pairs in three clusters.

These three clusters were formed mainly due to the similarity of the decision profile in re‐ gards to the cows' culling age in the production systems of Brasilândia and Montes Claros (8.2 and 8.5 years), Ribas Rio Pardo and Porangatu (9.2 and 9.0 years), Rio Verde and São Gabriel d'Oeste (10.3 and 10.7 years). In practice, however, the 5x5 grid didn't perform well, as it hadn't grouped other production systems with similar profiles. Thus, this grid didn't provide the basis for an accurate synthesis of convergence points among the other systems under assessment.

The 5x5 grid allocated many single municipalities in the clusters, i.e., one DMU in one clus‐ ter. Therefore, this topology was not suitable for this study. We also analyzed the 4x4 grid. We inferred that the 4x4 typology was also not adequate for this study as we obtained simi‐ lar results to the 5x5 grid.

**Figure 2.** DMUs' distribution for the 5x5 grid.

**Figure 1.** Cross-evaluation matrix. In the diagonal are the scores computed by the self-evaluations. Other cells show

 Alvorada TO #1 Amambai MS #2 Aquidauana MS #3 Bonito MS #4 Brasilândia MS #5 Montes Claros MG #12 Camapuã MS #6 Carlos Chagas MG #7 Catalão GO #8 Corumbá MS #9 Itamarajú BA #10 Lavras do Sul RS #11 Niquelândia GO #13 Paraíso do Tocantins TO #14 Porangatu GO #15 Ribas Rio Pardo MS #16 Rio Verde GO #17 São Gabriel d'Oeste MS #18 Tupã SP #19 Uberaba MG #20 Uberlândia MG #21

**DMUs Municipality State Code**

the peer-evaluations.

**Cluster**

78 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

**Table 4.** Clusters obtained with the 5x5 grid.

The set shown in Table 5 is from the 3x3 grid. We observe that some DMUs were in the same clusters they were allocated in the 5x5 grid. This was the case of the cluster composed of Brasilândia and Montes Claros. In this cluster were added the DMUs Catalão, Ribas Rio Pardo, Tupã and Uberlândia. The same occurred with the group Rio Verde – São Gabriel d'Oeste that in the 3x3 grid was also consisted of Niquelândia and Porangatu municipalities. We may point out that Porangatu and Ribas Rio Pardo were no longer in the same cluster, which was a consequence of this new grid.

profile of discarding cows and bulls. Bulls' disposal is directly related to purchase: the high‐

Ex-Post Clustering of Brazilian Beef Cattle Farms Using Soms and Cross-Evaluation Dea Models

http://dx.doi.org/10.5772/51324

81

By limiting the number of clusters to nine (3x3 grid) we obtained three clusters with only one municipality: cluster 1 (Alvorada), cluster 6 (Carlos Chagas) and cluster 7 (Itamarajú). These three municipalities had unique profiles of decision- making with respect to the cull‐ ing of cows and bulls. In Alvorada, cows and bulls culling rates were the lowest among the municipalities analyzed (8 and 13%, respectively). In Carlos Chagas, the rates were higher (18 and 21%). In Itamarajú, the cows disposal rate was small (12%), but the bulls disposal

The third grid used was the 2x2. The results are displayed in Table 6. Comparing these re‐ sults with the ones from the 3x3 grid we observe that most of the previous allocation re‐ mained unchanged and there were mergers according to the clusters similarities. The exception is cluster 8 in the 3x3 grid (Niquelândia, Porangatu, Rio Verde and São Gabriel d'Oeste): in the current configuration its DMUs were inserted into other clusters (Niquelân‐ dia and Porangatu are in cluster 3 along with Bonito, Brasilândia, Catalão, Itamarajú, Lavras do Sul, Montes Claros, Ribas Rio Pardo, Tupã and Uberlândia; Rio Verde and São Gabriel

Although these clusters were different from the ones obtained with the 3x3 grid, they were compatible with the results from 5x5 configuration, where the pairs Rio Verde – Gabriel

d'Oeste were in cluster 2 along with Aquidauana, Camapuã and Corumbá).

d'Oeste and Porangatu – Ribas Rio Pardo were allocated in different groups.

er is the annual discard, it increases the need for annual replacements.

rate was the highest among the production systems evaluated (24%).

Figure 3 shows the DMUs' distribution for the 3x3 dimension grid.

**Figure 3.** DMUs' distribution for the 3x3 grid.


**Table 5.** Clusters obtained with the 3x3 grid.

This situation was probably a result of the same decision-making profile of the municipali‐ ties that were grouped, in regards the dispose of cows and bulls in their production systems. The 3x3 grid provided balanced groups concerning the variables, especially in relation to the profile of discarding cows and bulls. Bulls' disposal is directly related to purchase: the high‐ er is the annual discard, it increases the need for annual replacements.

By limiting the number of clusters to nine (3x3 grid) we obtained three clusters with only one municipality: cluster 1 (Alvorada), cluster 6 (Carlos Chagas) and cluster 7 (Itamarajú). These three municipalities had unique profiles of decision- making with respect to the cull‐ ing of cows and bulls. In Alvorada, cows and bulls culling rates were the lowest among the municipalities analyzed (8 and 13%, respectively). In Carlos Chagas, the rates were higher (18 and 21%). In Itamarajú, the cows disposal rate was small (12%), but the bulls disposal rate was the highest among the production systems evaluated (24%).

Figure 3 shows the DMUs' distribution for the 3x3 dimension grid.

**Figure 3.** DMUs' distribution for the 3x3 grid.

d'Oeste that in the 3x3 grid was also consisted of Niquelândia and Porangatu municipalities. We may point out that Porangatu and Ribas Rio Pardo were no longer in the same cluster,

> Alvorada TO #1 Amambai MS #2 Paraíso do Tocantins TO #14 Aquidauana MS #3 Camapuã MS #6 Corumbá MS #9 Bonito MS #4 Lavras do Sul RS #11 Brasilândia MS #5 Catalão GO #8 Montes Claros MG #12 Ribas Rio Pardo MS #16 Tupã SP #19 Uberlândia MG #21 Carlos Chagas MG #7 Itamarajú BA #10 Niquelândia GO #13 Porangatu GO #15 Rio Verde GO #17 São Gabriel d'Oeste MS #18 Uberaba MG #20

This situation was probably a result of the same decision-making profile of the municipali‐ ties that were grouped, in regards the dispose of cows and bulls in their production systems. The 3x3 grid provided balanced groups concerning the variables, especially in relation to the

**DMUs Municipality State Code**

which was a consequence of this new grid.

80 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

**Table 5.** Clusters obtained with the 3x3 grid.

**Cluster**

The third grid used was the 2x2. The results are displayed in Table 6. Comparing these re‐ sults with the ones from the 3x3 grid we observe that most of the previous allocation re‐ mained unchanged and there were mergers according to the clusters similarities. The exception is cluster 8 in the 3x3 grid (Niquelândia, Porangatu, Rio Verde and São Gabriel d'Oeste): in the current configuration its DMUs were inserted into other clusters (Niquelân‐ dia and Porangatu are in cluster 3 along with Bonito, Brasilândia, Catalão, Itamarajú, Lavras do Sul, Montes Claros, Ribas Rio Pardo, Tupã and Uberlândia; Rio Verde and São Gabriel d'Oeste were in cluster 2 along with Aquidauana, Camapuã and Corumbá).

Although these clusters were different from the ones obtained with the 3x3 grid, they were compatible with the results from 5x5 configuration, where the pairs Rio Verde – Gabriel d'Oeste and Porangatu – Ribas Rio Pardo were allocated in different groups.


discard rates of cows and bulls of 10% and 13%, and 18% and 15%, respectively. This im‐ plies that the former group (with lower rates) probably had herds with a lower replacement dynamics. In the latter cluster, which had a lower rate of bulls' replacement, there was less

Ex-Post Clustering of Brazilian Beef Cattle Farms Using Soms and Cross-Evaluation Dea Models

http://dx.doi.org/10.5772/51324

83

Finally, the 2x1 dimension grid was used and the results are depicted in Table 7 and Figure

 Alvorada TO #1 Amambai MS #2 Aquidauana MS #3 Brasilândia MS #5 Camapuã MS #6 Catalão GO #8 Corumbá MS #9 Itamarajú BA #10 Montes Claros MG #12 Niquelândia GO #13

**DMUs Municipality State Code**

investment in improving the genetic quality of the herd.

**Figure 4.** DMUs' distribution for the 2x2 grid.

5 shows its DMUs' distribution.

**Cluster**

**Table 6.** Clusters obtained with the 2x2 grid.

The DMUs' allocation in regards to the 2x2 grid is shown in Figure 4. The biggest cluster had 11 municipalities: Bonito, Brasilândia, Catalão, Itamarajú, Lavras do Sul, Montes Claros, Niquelândia, Porangatu, Ribas Rio Pardo, Tupã and Uberlândia. In this group, cows and bulls discard rates had medium values (19 and 14%, respectively).

The smallest group was composed of two municipalities: Carlos Chagas and Uberaba. These municipalities exhibited higher rates for the discarding of cows and bulls (18% and 21%), which is a characteristic of production systems with a more dynamics profile, when com‐ pared to other clusters.

The clusters composed of three (Alvorada, Amambai and Paraíso do Tocantins) and five municipalities (Rio Verde, Aquidauana, Camapuã, São Gabriel d'Oeste and Corumbá) had discard rates of cows and bulls of 10% and 13%, and 18% and 15%, respectively. This im‐ plies that the former group (with lower rates) probably had herds with a lower replacement dynamics. In the latter cluster, which had a lower rate of bulls' replacement, there was less investment in improving the genetic quality of the herd.

**Figure 4.** DMUs' distribution for the 2x2 grid.

**Cluster**

82 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

**Table 6.** Clusters obtained with the 2x2 grid.

pared to other clusters.

**DMUs Municipality State Code**

 Alvorada TO #1 Amambai MS #2 Paraíso do Tocantins TO #14 Aquidauana MS #3 Camapuã MS #6 Corumbá MS #9 Rio Verde GO #17 São Gabriel d'Oeste MS #18 Bonito MS #4 Brasilândia MS #5 Catalão GO #8 Itamarajú BA #10 Lavras do Sul RS #11 Montes Claros MG #12 Niquelândia GO #13 Porangatu GO #15 Ribas Rio Pardo MS #16 Tupã SP #19 Uberlândia MG #21 Carlos Chagas MG #7 Uberaba MG #20

The DMUs' allocation in regards to the 2x2 grid is shown in Figure 4. The biggest cluster had 11 municipalities: Bonito, Brasilândia, Catalão, Itamarajú, Lavras do Sul, Montes Claros, Niquelândia, Porangatu, Ribas Rio Pardo, Tupã and Uberlândia. In this group, cows and

The smallest group was composed of two municipalities: Carlos Chagas and Uberaba. These municipalities exhibited higher rates for the discarding of cows and bulls (18% and 21%), which is a characteristic of production systems with a more dynamics profile, when com‐

The clusters composed of three (Alvorada, Amambai and Paraíso do Tocantins) and five municipalities (Rio Verde, Aquidauana, Camapuã, São Gabriel d'Oeste and Corumbá) had

bulls discard rates had medium values (19 and 14%, respectively).

Finally, the 2x1 dimension grid was used and the results are depicted in Table 7 and Figure 5 shows its DMUs' distribution.



ter 1. This was a result of the new grid used. Both groups had no outstanding difference be‐ tween the culling rates of cows and bulls (14 and 17% in cluster 1, 14 and 18% in cluster 2).

Ex-Post Clustering of Brazilian Beef Cattle Farms Using Soms and Cross-Evaluation Dea Models

http://dx.doi.org/10.5772/51324

85

This grid was not satisfactory to support the analysis of similarities and dissimilarities of the

In this paper we used Data Envelopment Analysis with SOMS to cluster the 21 beef cattle modal production systems into homogeneous groups according to their efficiency profiles. The efficiency profiles are derived from the cross-evaluation DEA approach. This approach allows a peer-evaluation instead of only considering the self-evaluation performed by the

The approach used here is different from the previous ones in the literature, in which SOMS are used to group the DMUs into homogeneous sets and then use DEA models to evaluate the DMUs belonging to each cluster. In our approach rather than performing ex-ante clus‐ tering we carried out ex-post grouping. We used the DEA cross-efficiencies as inputs to the SOMS procedure. It is important to stress that we didn't use the classic DEA efficiency scores, but the efficiency profiles of each DMU derived from the cross-evaluation matrix.

The proposed approach yielded interesting results for 3x3 and 2x2 grids. In using these grids, the production systems developed in the 21 assessed municipalities were grouped, particularly, in relation to the inlet-outlet dynamics of important animals' categories of the breeding herd. The different decision-making profiles of culling of cows and bulls, and con‐ sequently of the purchase of bulls, were satisfactorily grouped relative to the efficiency pro‐

To National Council for Scientific and Technological Development (CNPq) for the finan‐

, Urbano Gomes Pinto de Abreu4

,

,

João Carlos Correia Baptista Soares de Mello1\*, Eliane Gonçalves Gomes2

and Sergio de Zen5

, Luiz Biondi Neto3

\*Address all correspondence to: jcsmello@producao.uff.br

beef cattle production systems profiles evaluated.

**7. Conclusions**

classic DEA models.

file observed in each production system.

**Acknowledgements**

cial support.

**Author details**

Lidia Angulo Meza1

Thiago Bernardino de Carvalho5

1 UFF. Passo da Pátria 156, Brazil

**Table 7.** Clusters obtained with the 2x1 grid.

We notice that a small cluster was formed with Bonito, Carlos Chagas, Lavras do Sul and Uberaba. Bonito and Lavras do Sul were part of one cluster in the 3x3 grid and in the 2x2 grid they were included in cluster 3 along with Brasilândia, Catalão, Itamarajú, Montes Claros, Niquelândia, Porangatu, Ribas Rio Pardo, Tupã and Uberlândia, which were in clus‐ ter 1. This was a result of the new grid used. Both groups had no outstanding difference be‐ tween the culling rates of cows and bulls (14 and 17% in cluster 1, 14 and 18% in cluster 2).

This grid was not satisfactory to support the analysis of similarities and dissimilarities of the beef cattle production systems profiles evaluated.

## **7. Conclusions**

**Cluster**

84 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

**Table 7.** Clusters obtained with the 2x1 grid.

**Figure 5.** DMUs' distribution for the 2x1 grid.

**DMUs Municipality State Code**

 Paraíso do Tocantins TO #14 Porangatu GO #15 Ribas Rio Pardo MS #16 Rio Verde GO #17 São Gabriel d'Oeste MS #18 Tupã SP #19 Uberlândia MG #21 Bonito MS #4 Carlos Chagas MG #7 Lavras do Sul RS #11 Uberaba MG #20

We notice that a small cluster was formed with Bonito, Carlos Chagas, Lavras do Sul and Uberaba. Bonito and Lavras do Sul were part of one cluster in the 3x3 grid and in the 2x2 grid they were included in cluster 3 along with Brasilândia, Catalão, Itamarajú, Montes Claros, Niquelândia, Porangatu, Ribas Rio Pardo, Tupã and Uberlândia, which were in clus‐ In this paper we used Data Envelopment Analysis with SOMS to cluster the 21 beef cattle modal production systems into homogeneous groups according to their efficiency profiles. The efficiency profiles are derived from the cross-evaluation DEA approach. This approach allows a peer-evaluation instead of only considering the self-evaluation performed by the classic DEA models.

The approach used here is different from the previous ones in the literature, in which SOMS are used to group the DMUs into homogeneous sets and then use DEA models to evaluate the DMUs belonging to each cluster. In our approach rather than performing ex-ante clus‐ tering we carried out ex-post grouping. We used the DEA cross-efficiencies as inputs to the SOMS procedure. It is important to stress that we didn't use the classic DEA efficiency scores, but the efficiency profiles of each DMU derived from the cross-evaluation matrix.

The proposed approach yielded interesting results for 3x3 and 2x2 grids. In using these grids, the production systems developed in the 21 assessed municipalities were grouped, particularly, in relation to the inlet-outlet dynamics of important animals' categories of the breeding herd. The different decision-making profiles of culling of cows and bulls, and con‐ sequently of the purchase of bulls, were satisfactorily grouped relative to the efficiency pro‐ file observed in each production system.

## **Acknowledgements**

To National Council for Scientific and Technological Development (CNPq) for the finan‐ cial support.

## **Author details**

João Carlos Correia Baptista Soares de Mello1\*, Eliane Gonçalves Gomes2 , Lidia Angulo Meza1 , Luiz Biondi Neto3 , Urbano Gomes Pinto de Abreu4 , Thiago Bernardino de Carvalho5 and Sergio de Zen5

\*Address all correspondence to: jcsmello@producao.uff.br

1 UFF. Passo da Pátria 156, Brazil

2 Brazilian Agricultural Research Corporation (Embrapa), Brazil

3 UERJ. Rua São Francisco Xavier 524, Brazil

4 Embrapa Pantanal. Caixa Postal 109, Brazil

5 Center for Advanced Studies on Applied Economics (CEPEA/ESALQ/USP), Brazil

### **References**

[1] Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of deci‐ sion-making units. *European Journal of Operational Research*, 2, 429-44.

[10] Sexton, T. R., Silkman, R. H., & Logan, A. J. (1986). Data Envelopment Analysis: Cri‐ tique and extensions. In: Silkman H, ed. Measuring efficiency: An assessment of data

Ex-Post Clustering of Brazilian Beef Cattle Farms Using Soms and Cross-Evaluation Dea Models

http://dx.doi.org/10.5772/51324

87

[11] Banker, R. D., Charnes, A., & Cooper, W. W. (1984). Some models for estimating tech‐ nical scale inefficiencies in data envelopment analysis. *Mngmt Sc.*, 30(9), 1078-92.

[12] Doyle, J. R., & Green, R. H. (1994). Efficiency and cross-efficiency in DEA derivations,

[13] Doyle, J. R., & Green, R. H. (1995). Cross-Evaluation in DEA: Improving Discrimina‐ tion Among DMUs. *Information Systems and Operational Research*, 33(3), 205-22.

[14] Appa, G., Argyris, N., & Williams, H. P. (2006). A methodology for cross-evaluation in DEA. Working Paper: London School of Economics and Political Science Novem‐

[15] Soares de, Mello. J. C. C. B., Lins, M. P. E., & Gomes, E. G. (2002). Construction of a

[16] Wu, J., Liang, L., & Chen, Y. (2009). DEA game cross-efficiency approach to Olympic

[17] Gomes, E. G., Soares de Mello, J. C. C. B., Angulo-Meza, L., Silveira , J. Q., Biondi Ne‐ to, L., & Abreu, U. G. P. d. (2012). Some Remarks About Negative Efficiencies in DEA Models. In: Holtzman Y, ed. Advanced Topics in Applied Operations Management.

[18] Liang, L., Wu, J., Cook, W. D., & Zhu, J. (2008). The DEA game cross-efficiency model

[19] Liang, L., Wu, J., Cook, W. D., & Zhu, J. (2008). Alternative secondary goals in DEA cross-efficiency evaluation. *International Journal of Production Economics*, 113(2),

[20] Mitra, P., Murthy, C. A., & Pal, S. K. (2002). Unsupervised feature selection using fea‐ ture similarity. *IEEE Transactions on Pattern Analysis and Machine Intelligence*, 24(3),

[21] Willshaw, D. J., Buneman, O. P., & Longuet-Higgins, H. C. (1969). Non-holographic

[22] Willshaw, D. J., & Von der, Malsburg. C. (1976). How patterned neural connections can be set up by self-organization. *Proceedings of the Royal Society of London Series B*,

[24] Haykin, S. (1999). Neural networks: a comprehensive foundation. nd ed. ed. New Jer‐

[23] Kohonen, T. (2001). Self-organizing maps. rd ed. ed. Berlin: Springer-Verlag.

smoothed DEA frontier. *Pesquisa Operacional.*, 28(2), 183-201.

and its nash equilibrium. *Operations Research*, 56(5), 1278-88.

meanings and uses. *Journal of the Operational Research Society*, 45, 567-78.

envelopment analysis. San Francisco Jossey-BassEditor: , 73-105.

ber 7.

rankings. *Omega*, 37(4), 909-18.

associative memory. *Nature*, 222, 960-2.

Rijeka InTech , 113-32.

1025-30.

301-12.

194, 431-45.

sey Prentice Hall


[10] Sexton, T. R., Silkman, R. H., & Logan, A. J. (1986). Data Envelopment Analysis: Cri‐ tique and extensions. In: Silkman H, ed. Measuring efficiency: An assessment of data envelopment analysis. San Francisco Jossey-BassEditor: , 73-105.

2 Brazilian Agricultural Research Corporation (Embrapa), Brazil

5 Center for Advanced Studies on Applied Economics (CEPEA/ESALQ/USP), Brazil

sion-making units. *European Journal of Operational Research*, 2, 429-44.

Neural Networks. *European Journal of Operational Research*, 206(2), 479-87.

[1] Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of deci‐

[2] Samoilenko, S., & Osei-Bryson, K. M. (2010). Determining sources of relative ineffi‐ ciency in heterogeneous samples: Methodology using Cluster Analysis, DEA and

[3] Sharma, M. J., & Yu, S. J. (2009). Performance based stratification and clustering for benchmarking of container terminals. Expert Systems with Applications PART 1) ,

[4] Churilov, L., & Flitman, A. (2006). Towards fair ranking of olympics achievements: The case of Sydney 2000. *Computers and Operations Research*, 33(7), 2057-2082.

[5] Biondi, Neto. L., Lins, M. P. E., Gomes, E. G., Soares de Mello., J. C. C. B., & Oliveira, F. S. (2004). Neural data envelopment analysis: a simulation. *International Journal of*

[6] Emrouznejad, A., & Shale, E. (2009). A combined neural network and DEA for meas‐ uring efficiency of large scale datasets. *Computers and Industrial Engineering*, 56(1),

[7] Çelebi, D., & Bayraktar, D. (2008). An integrated neural network and data envelop‐ ment analysis for supplier evaluation under incomplete information. *Expert Systems*

[8] Angulo-Meza, L., Biondi Neto., L., Brandão, L. C., Andrade, F. V. S., Soares de Mello, J. C. C. B., & Coelho, P. H. G. (2011). Modelling with self-organising maps and data envelopment analysis: A case study in educational evaluation. *Self organizing maps,*

[9] Paschoalino, F. F., & Soares de, Mello. J. C. C. B., Angulo-Meza L., Biondi Neto L. (2011). DMU clustering based on Cross Efficiency Evaluation. International Confer‐ ence on Data Envelopment Analysis and Its Applications to Management. Lima (Pe‐

3 UERJ. Rua São Francisco Xavier 524, Brazil

86 Developments and Applications of Self-Organizing Maps Applications of Self-Organizing Maps

4 Embrapa Pantanal. Caixa Postal 109, Brazil

**References**

5016-5022.

249-54.

ru).

*Industrial Engineering*, 11, 14-24.

*with Applications*, 35(4), 1698-710.

*new achievements*, Vienna, Intech, 71-88.


[25] Bishop, C. M. (1995). Neural networks for pattern recognition. New York, Oxford University Press.

**Chapter 5**

**A Self – Organizing Map Based Strategy for**

Even though education and knowledge are processes inherent to human development and are present in all cultures since the earliest and most remote age, the educational model adopted today had its origins in Ancient Greece, where the first signs of appreciation of cul‐ tural knowledge took place [1]. At the time, the most traditional way to prepare young indi‐ viduals of any social class to social integration was through individual processes of teaching, whether in daily life, with their parents, or in contact with masters and artisans. Although, the more privileged classes enjoyed other types of learning, such as access to reading, writing and other areas of knowledge, this process was always conducted on an in‐

The need to generalize the teaching of reading, writing and the so-called general culture among the less privileged social strata caused an increase in the number of students in rela‐ tion to the number of teachers available. This fact prompted educators that time to seek a teaching model that could bring knowledge from the educators to a maximum number of individuals at the same time. Given this need, the Greeks developed the earliest forms of

The school and the way students are organized in the classroom have also undergone vari‐ ous transformations throughout history [2]. Initially, they were organized in large groups in a single classroom, and guided by a teacher or tutor who had different concepts that he judged to be of common interest, combined with specific content, targeted to smaller groups or individual students. Later, new forms of organization of schools and classrooms emerged, such as the structuring of the content presented according to age, the division of

> © 2012 Silva et al.; licensee InTech. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,

© 2012 Silva et al.; licensee InTech. This is a paper distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

distribution, and reproduction in any medium, provided the original work is properly cited.

grouping students in order to maximize their teaching activities [2].

**Heterogeneous Teaming**

Huliane M. Silva, Cícero A. Silva and

Additional information is available at the end of the chapter

Flavius L. Gorgônio

**1. Introduction**

dividual basis.

http://dx.doi.org/10.5772/52776


## **Chapter 5**
