Implementation and Comparison Clustering Algorithms with Duplicate Entities Detection Purpose in Data Bases

The aim of study is finding appropriate clustering algorithms for iteration detection issues on existing data set. The issue of identifying iterative records issue is one of the challenging issues in the field of databases. As a result, finding appropriate algorithms in this field helps significantly to organize information and extract the correct answer from different queries of database. This study is a combination of the author's previous studies. In this study, 4 algorithms, K-Means, Single-Linkage, DBSCAN and Self-Organizing Maps have been implemented and compared. F1 measure was used in order to evaluate precision and quality of clustering that by evaluating the obtained results, the SOM algorithm obtained high accuracy. However, the base SOM algorithm due to using Euclidean distance has some defects in solving real problems. In order to solve these defects, Gaussian kernel has been used to measure Euclidean distance that by studying obtained results it was seen that KSOM algorithm has higher F1 measure than base SOM algorithm. Initializing weight vector in SOM algorithm is one of the main and effective problems in convergence of algorithm. In this research we presented a method that optimized initialing weight step. Presented method reduces the number of iteration in comparison than basic method as it increases the run rate.

並列關鍵字

Clustering ； DBSCAN ； F1 measure ； K-means ； self-organizing maps ； single-linkage

國際替代計量

全文下載

主題瀏覽