**3. Methodology**

This study will investigate the impact of three leadership styles (laissez faire, transactional and transformational) on the willingness of employees to exert extra effort, effectiveness and satisfaction. In this study, the researchers have opted for a Multifactor Leadership Questionnaire (MLQ) for collecting data. MLQ is a popular instrument for the study of leadership styles [24]. MLQ is a proven instrument for measuring the leadership styles since its inception in 1985. The study has employed a five-point Likert scale questionnaire to collect quantitative data (where 1 being strongly disagree and 5 being strongly agree). The data were collected from the employees of small- to medium-sized enterprises (SMEs) in the Bay of Plenty Region of New Zealand. A total of 24 organisations were selected to gather data samples (within the range of 200–225) for this research. The participants answered questionnaires that portray the style of leadership of their immediate level manager. The goal of this research was to find and compare leadership styles of managers and their implications from the employee's point of view.

effectiveness and leadership outcome will be mapped to low, medium and high categorical values. The low category covers scores 1 to 3, medium between 3 and 4 and high from 4 to 5.

**Table 1.** Frequency distribution of extra effort, effectiveness, satisfaction and (leadership) outcome after performing data

Investigation on the Impact of Leadership Styles Using Data Mining Techniques

http://dx.doi.org/10.5772/intechopen.78660

145

**Labels/measure Extra effort Effectiveness Satisfaction Outcome** Low 57 33 42 33 Medium 102 102 94 94 High 43 67 66 75

A total of 210 samples were collected, of which three were discarded as these had missing values and another five were removed after conducting outlier analysis. The final 202 samples were used to analyse the data. The reliability test was carried out to demonstrate the internal consistency of various measures used in this research. Coefficient alpha or commonly known as Cronbach's alpha was used to test the reliability of data, and the values of above 80% were

The total number of participants surveyed was 202 of which 66% were females and 34% were males. Approximately half of the participants (51%) were 25 to 34 years of age. The participants between the age 18–24 and 35–44 were 16% each. In terms of education, 36% of the participants had completed their bachelor's degree. There were 26 and 24% of participants with diploma and postgraduate qualifications. The remaining 14% had their higher secondary school completed. Participant's prior employment statistics showed that 44% of them had more than 7 years of total experience followed by 29% of them between 1 and 3 years of work experience. Participants with less than 1 year of experience and experience between 5 and 7 years were 7 and 8%, respectively. This depicts that the participants had fairly good experience of working in organisations. Working at the current job, more than half (52%) of the participants fall under the category of below 1 year of working at the current organisation, followed by 20% for 1–3 years of working at current job. Industries covered in this study ranges from health, hospitality, customer service, education and trades with health segment employing more females stood top at 34%, followed by hospitality at 25% and customer service 21%.

**Table 2** presents pairwise correlation coefficients and the statistical significance of each coefficient showing the relationships between leadership styles and employee outcomes. The

The frequency table of all class labels are mentioned in **Table 1**.

**4. Experimental results**

transformation.

**4.1. Descriptive analysis**

obtained in all factors of this questionnaire.

**4.2. Correlation and regression analysis**

A total of 18 questions were designed to record the feedback on independent variables. Six questions each were designed to gather feedback on each leadership styles (transformational, transactional and laissez faire). The responses of these six questions were accumulated to find the score of each leadership style. Another nine questions were designed to gather responses of dependent variables. Three questions each for extra effort, effectiveness and satisfaction were designed, and the responses were calculated by taking the mean of each category questions. The pooled variable 'leadership outcome' was derived by taking the mean of extra effort, effectiveness and satisfaction.

There are number of data mining algorithms available in the literature such as artificial neural networks, support vector machines, decision trees, rule-based algorithms, k-nearest neighbours and Bayesian algorithms [17]. However, in this research, three algorithms, namely OneR, J48, and Modlem are used. These algorithms are used in this research as the outcome of these algorithms are in the form of a decision tree or a set of rules, which are easy to interpret.

OneR stands for One Rule or one level decision tree. This algorithm processes all input variables and selected single input variable that best model (classify) the output variable [30]. This is a very simple algorithm but has accuracy comparable to the state-of-the-art data mining algorithms. Mostly, OneR algorithm is used as a baseline algorithm against which the performance of other algorithms can be compared. Here, this algorithm will also be used as a baseline algorithm. The model accuracy and the degree of information extracted from the data will be compared to other data mining algorithms.

J48 or C4.5 is a decision tree-based classification algorithm proposed by Quinlin [31]. J48 builds the decision tree on the training data using the concept of information entropy. At each node, the algorithm selects a variable that split the data into defined classes with highest information gain. The variable with highest information gain is select to expand the node. Modlem algorithm generates rules based on rough set theory and more information can be found in [32].

In order to run data mining algorithms, the dependent variables' score will be transformed into categorical variables. The numeric scores recorded on the extra effort, satisfaction,


**Table 1.** Frequency distribution of extra effort, effectiveness, satisfaction and (leadership) outcome after performing data transformation.

effectiveness and leadership outcome will be mapped to low, medium and high categorical values. The low category covers scores 1 to 3, medium between 3 and 4 and high from 4 to 5. The frequency table of all class labels are mentioned in **Table 1**.
