Classification of microarray gene expression data using associative classification

Abstract

Data mining is one of the interdisciplinary fields on the research aspect. Recently, association rules have become an important concept for classification purposes, called Associative Classification. It is a data mining technique that combines association rule mining and classification technique to build classification models. The core objective of this work is to classify the gene expression using Associative Classification algorithm. Our research work has been formulated based on association rule and classification mining. The proposed algorithm contains four phases called as Statistical Gene filtering, Discretization, Class Association Rules and prediction or class assignment. The gene filtering phase is to find the differentially expressed genes and select the significant genes in the specific gene expression. The discretization phase is to convert the continuous values into discrete values and substitute the gene values into gene intervals. The role of the Class Association Rule phase is to generate the set of class association rules using closed frequent itemset and generate the classifier model. The last phase predicts the class from trained model using scoring function. The experimental results carried out by using breast cancer gene expression data which are available on the NCBI online biological database. It has been tested with two class and multi-class datasets and compared with the classical classification algorithms such as Linear Discriminant Analysis, SVM, and Decision Tree. The performance of the classifier model is evaluated using Leave-One-Out-Cross Validation method. The result of this work is used to the drug designer for the pathway analysis and disease treatment decisions.