The increase in the number of documents has aggravated the difficulty of classifying those documents according to specific needs. Clustering analysis in a distributed environment is a thrust area in artificial intelligence and data mining. Its fundamental task is to utilize characters to compute the degree of related corresponding relationship between objects and to accomplish automatic classification without earlier knowledge. Document clustering utilizes clustering technique to gather the documents of high resemblance collectively by computing the documents resemblance. Recent studies have shown that ontologies are useful in improving the performance of document clustering. Ontology is concerned with the conceptualization of a domain into an individual identifiable format and machine-readable format containing entities, attributes, relationships, and axioms. By analyzing types of techniques for document clustering, a better clustering technique depending on Genetic Algorithm (GA) is determined. Non-Dominated Ranked Genetic Algorithm (NRGA) is used in this paper for clustering, which has the capability of providing a better classification result. The experiment is conducted in 20 newsgroups data set for evaluating the proposed technique. The result shows that the proposed approach is very effective in clustering the documents in the distributed environment.
TopIntroduction
Clustering of document is not only an efficient way of organizing, summarizing, extracting and retrieving of information but also an important method in documentation. Initially, clustering is applied for enhancing the precision or in recalling information retrieval techniques. Clustering analysis is an important field of research intelligence and data mining. The fundamental concept of clustering analysis springs from characteristic to calculate the degree of similar relationship among objects and to attain automatic classification. Recently, clustering data in large data sets (Sharma et al., 2009) has gained a lot of attention in myriad research domains.
Moreover, clustering technique is applicable in certain areas which involve browsing gathered data or in categorizing the outcome, provided by the search engine in terms of replies to the queries provided by the user. Document clustering can also be applied in producing the hierarchical grouping of document. In order to search and retrieve the information efficiently in Document Management Systems (DMSs), the metadata set should be created for the documents with enough details. One metadata per se is not enough for the document management systems. This is because the various document types need various attributes for distinguishing appropriately.
Though there are various document clustering techniques available in the literature but the existing clustering algorithms suffer from problems of practical applicability. The accuracy of the existing clustering approaches is a major concern. The time taken for active clustering of documents is more in large databases. So, a novel technique is needed for effective document clustering with very high accuracy but less complexity.
The discrete use of importance of using the ontology (Grootjen & Weide, 2005) increased in document clustering scheme as this technique seems to the accuracy of clustering. Whereas ontologies provide a position of legitimacy, research in this field leads to superior importance in the arising of different disputes featured in the modern digital situation. With the intention of providing efficacious solutions to the various drawbacks related to present search techniques, ontologies are widely implemented for creating viable document clustering techniques. An example of research concern in the field of ontology (Leukel & Sugumaran, 2009) relates to the Google which serves as search engine for semantic web documents, terms and data found on the internet. For retrieving information from the complex characteristic of the digital information available in digital libraries, ontologies can be successfully used for producing an efficient information retrieval system.
Ontologies (Murthy & Chowdhury, 1996) are introduced as modeling technology for structured metadata definition within the document clustering system. With the obtained metadata the clustering can be performed using the Genetic Algorithms (GAs) (Koller et al., 1997). GA is a search technique that is based on natural genetic and selection merging the idea of survival of the fittest with structured interchanges. Genetic algorithm (Sun et al., 2008) is a famous technique to deal with complex search problems by implementing an evolutionary stochastic search because genetic algorithm can be very effectively applied to various challenging optimization problems. The NP-hard nature of the clustering technique makes genetic algorithm a natural choice for solving it. A common objective function in these implementations is decreasing the square error.