Pattern Discovery using Fuzzy FP-growth Algorithm from Gene Expression Data

Abstract

The goal of microarray experiments is to identify genes that are differentially transcribed with respect to different biological conditions of cell cultures and samples. Hence, method of data analysis needs to be carefully evaluated such as clustering, classification, prediction etc. In this paper, we have proposed an efficient frequent pattern based clustering to find the gene which forms frequent patterns showing similar phenotypes leading to specific symptoms for specific disease. In past, most of the approaches for finding frequent patterns were based on Apriori algorithm, which generates and tests candidate itemsets (gene sets) level by level. This processing causes iterative database (dataset) scans and high computational costs. Apriori algorithm also suffers from mapping the support and confidence framework to a crisp boundary. Our hybridized Fuzzy FP-growth approach not only outperforms the Apriori with respect to computational costs, but also it builds a tight tree structure to keep the membership values of fuzzy region to overcome the sharp boundary problem and it also takes care of scalability issues as the number of genes and condition increases.