2. Related work

extraordinary, complex, and fascinating natural phenomena. Of particular interest is the intelligent behavior of some animal species to find a solution to solve a problem and maintain the perfect balance of the environment surrounding.

This is the main idea of computation inspired by nature, that is, to imitate the steps that nature has developed and adapt them to find a solution of a problem, thus converting it into a bio-inspired algorithm. This is the main reason for the implementation of an algorithm that exploits the collective intelligence of a Bee Colony as an alternative for clustering. This chapter presents an innovative approach for Web services clustering using a hybrid algorithm based on the Artificial Bee Colony, K-

The work described in this chapter is part of a research project whose objective is to design and implement a semantic directory of Web services using efficient, fully automated methods to allow the organization, composition, and classification of

Figure 1 shows the general architecture of the semantic directory and a methodology to construct and manage the directory. The methodology consists of the

1. Public Web service retrieval aims at searching over the Internet to find and copy Web service descriptions (files formatted in WSDL description language) in a

local file directory. Web service retrieval is executed through crawlers designed specifically to parse and identify links to files in WSDL.

2. Extraction and analysis of Web services consists of parsing every Web service description file and the extraction of specific data, such as: method names, input and output parameters, and port types to facilitate the automatic invocation. Extracted data is stored in an ontology model. For this phase, we use a tool which transforms WSDL files into an ontological representation.

3.Web service similarity calculation is an important phase of the methodology, because classification and clustering of Web services requires the calculation of

distances between services. In order to calculate similarities, different measures can be implemented and combined to obtain better results.

means, and Consensus.

following phases:

Figure 1.

10

Semantic directory of Web services.

Web services with a semantic approach.

Advanced Analytics and Artificial Intelligence Applications

Web service classification and clustering is a topic addressed form different perspectives such as statistical, stochastic, and novel approaches based on bio-inspired algorithms. Additionally, semantic approaches to describe, discover, and invoke Web services have been studied to propose novel clustering algorithms. In this section a revision of related work is presented considering two trends: reported work that address clustering and classification of Web services and clustering approaches based on bio-inspired algorithms. Table 1 presents the main characteristics of related work.

In 2009, Liang et al. [1] proposed a method for Web service categorization considering keywords and semantic relations between elements of the description. Their proposed methodology involves preprocessing WSDL documents, rough clustering by labeling Web services with class tag, and fine clustering.

In 2009, Platzer et al. [2] described a scalable approach for clustering very large service repositories. They use a statistical clustering algorithm enhancing a vector space to support the search of services related to a given query.

In 2012, Pop et al. [3] presented two approaches for service clustering, one inspired by the behavior of the birds and other inspired by the behavior of ants. They implemented methods to evaluate the semantic similarity between services.

In 2013, Du et al. [4] presented an approach for clustering Web services based on functional similarity and refinement of clusters using a concept position vector.

In 2014, Wu et al. [5] presented an approach which consists of three modules: data preprocessing, Web service tag recommendation, and Web services clustering. The first module consists of building a content vector formed with nouns, verbs, or adjectives. Authors use different features and different approaches for similarity computation. For content use the normalized Google distance, for data types and messages they use a basic match similarity, and for tag similarity they apply the Jaccard coefficient.

In 2014, Prakash and Singh [6] compared the performance of evolutionary algorithms: Genetic Algorithm, Differential Evolution, Particle Swarm Optimization, and Artificial Bee Colony for clustering three real and one synthetic data sets.

In 2017, Sahoo [7] presented a two-step ABC algorithm for data clustering problems. Authors addressed the three problems of the ABC algorithm such as initial positions of food sources, solution search equation, and abandoned food location.

In 2018, Kotekar and Kamath [8] described a Web services clustering approach based on Cat Swarm Optimization (CSO), which emulates the social behavior of cats in nature.

Automated Web services clustering is useful to facilitate service search, service discovery, service composition, and service substitution. Of particular interest is the representation of Web services through ontologies because the purpose of this work is the automatic organization of any collection (public or private) of Web service in ontologies and their semantic enrichment by classification and clustering.


Work

13

Singh

one synthetic

Optimization,

 and Artificial

[6]

Sahoo

Data is

Two-step ABC algorithm

 No

 Euclidean distance

 Not for

Data sets are

The clustering approach is based

The clustering method is not

applied to Web services.

Similarity measure is not

semantic

Bio-Inspired Hybrid Algorithm for Web Services Clustering

DOI: http://dx.doi.org/10.5772/intechopen.85200

Similarity measure is not

semantic

Web

downloaded

on novel bio-inspired

algorithms

services

from the UCI

repository

[7]

downloaded

from UCI repository

Kotekar

WSDL

Cat Swarm

Optimization

No

 Euclidean distance

1083

 OWL-TC4

The clustering approach is based

collection

on novel bio-inspired

algorithms

and TF-IDF

and

documents

(CSO)

Kamath

extracted from

[8]

Table 1. Comparison

 of related work.

OWL-TC4

data sets

Bee Colony

 Input data

 Clustering approach

 Use of

Similarity

Number

Service

Benefits

Limitations

of Web

repository

services

used

ontologies

approach


Table 1. Comparisonofrelated

 work.
