Genetic Algorithm Based Approach to Data Mining

Abstract

In this thesis four different scenarios are addressed where we try to generate optimistic solutions to Data Mining issues. First without the use of Genetic algorithms in acquiring the best pageusingaQuerytermsynonymcombination. InthesecondscenarioweuseaGeneticAlgorithm based approach to a data mining related issue of generating the best rules using genetic algorithm from rules generated by Apriori, the popular association rule mining algorithm. In the third scenario we provide recommendations using an algorithm of association rule mining namely Eclat and followed by using genetic algorithm to optimise the rules generated and and#64257;nally in the fourth scenario the speeding up of rule generation is handled, using genetic algorithm variants in parallel and hierarchical architectures. In all cases we are handling large data sets. Generating relevant information from the web for a user has become a question of great concern. Search engines return pages depending on a Ranking algorithm based on links, to and from the page and on how popular a page is, with respect to the hits received by users. In most casesthepagesreturnedaretoomanyandirrelevant. Asystemassociatedwithusingsynonyms of the words(terms) of a query given by a user has proved to be useful. A combination of such synonyms and#64257;red to a search engine has returned relevant information pages. The pages are examined for relevance to speciand#64257;c users and usefulness of content to a speciand#64257;c domain. The pages are also examined for their positions using ranking tools, trustworthiness tools and intent drifting. It is found that the pages returned using the method of combining synonyms of terms of the user query are placed at better ranking positions. Finding interesting and optimistic rules (best rules) from a dataset, which are used to normallyrepresentthedatasetmaximallybecomesdifand#64257;cultwhendealingwithdatarepositories. Associationrulescancatertoclassiand#64257;cationofthisdataandtargetscanbesetasperrequirement of a user. These association rules and#64257;nd all rules. These rules are actionable in