Sequence Gene Prediction based on Association Rule Mining

Gene expression databases are increasing day by day with the help of Microarray and SAGE expression profiling techniques. Prediction of disease-causing genes is a fundamental challenge on these databases. This paper deals with a supervised learning approach in which genes are revealed sequentially and the goal is to determine what will be next revealed in the gene expression database. Here we consider the association rule mining as a predictive tool. In association rule mining, rules are sorted based on the firm minimum support. When we used firm minimum support in gene expression database important rules were eliminated. To obviate this problem we present a balanced confidence measure which provides the opportunity of selecting perfect rules that have emerged least times in the training circle. Experiments on 4 microarray datasets show that the balanced confidence algorithm outperforms on rule extraction.