An Integrated Framework for Mixed Data Clustering Using Self Organizing Map | Science Publications

Journal of Computer Science

Hari P. Devaraj¹  M. Punithavalli¹ 

关键词: Attribute-oriented induction;    clustering technique;    data mining;    training pattern;    self-organizing map;    batch learning;    Better Matching Unit (BMU);    numeric attributes;    scientific data analysis;   

DOI  :  10.3844/jcssp.2011.1639.1645

学科分类：计算机科学（综合）

来源: Science Publications

【 摘 要 】

Problem statement: Clustering plays an important role in data mining of large data andhelps in analysis. This develops a vast importance in research field for providing better clusteringtechnique. There are several techniques exists for clustering the similar kind of data. But only veryfew techniques exists for clustering mixed data items. This leads to the requirement of betterclustering technique for classification of mixed data. The cluster must be such that the similarity ofitems within the clusters is increased and the similarity of items from different clusters must bereduced. The existing techniques possess several advantages and at the same time variousdisadvantages also exists. Approach: To overcome those drawbacks, Self-Organizing Map (SOM)and Extended Attribute-Oriented Induction (EAOI) for clustering mixed data type data can be used.This will take more time for clustering. A modified SOM was proposed based on batch learning.Results: The experimentation for the proposed technique was carried with the help of UCI AdultData Set. The number of clusters resulted for the proposed technique is lesser when compared to theusage of SOM. Also the outliers were not obtained by using the proposed technique. Conclusion:The experimental suggests that the proposed technique can be used to cluster the mixed data itemswith better accuracy of classification.

【 授权许可】

Unknown