Optimized defect prediction model using statistical process control and Correlation-Based feature selection method

Abstract

Defects are the flaws in software development process that causes the software to perform in an unexpected manner and produce erroneous outputs. Detecting these defects is an important task to ensure the quality of the software product. Defect prediction models acts as quality indicators that helps in detecting the defective components in the early phases of software development cycle. These models leads to reduced rework effort, more stable products and improved customer satisfaction. It is hard to find the high risk components that are major contributors for the defects from large number of variables. Thus feature selection is a very important aspect associated with defect analysis. Here we propose a defect prediction model to control the quality of software products using statistical process control. The key contributors for building the prediction models are derived using Correlation and ANOVA based feature selection methods. The proposed model is evaluated using benchmark dataset and the results are promising when compared with standard classification models.