An automated software failure prediction technique using hybrid machine learning algorithms

Abstract

Many sophisticated applications have been emerged in various industries due to the rapid growth of software technologies. Especially, the business organizations utilize the services of software-based applications to provide a state-of-the-art service. However, fault prediction in a software is a biggest challenge that needs to be addressed by the industries to improve the growth of their business. Therefore, there is a need for new techniques to perform fault prediction at an early stage of software life cycle so that software defects can be avoided in later stage. To overcome the issues in manual prediction, many prediction techniques are available that can predict the defects automatically. All of the available techniques are based on the pattern learning that finds the fault in the software based on the previously learned similar patterns. Even though many fault findings techniques are available, still there are some challenges to achieve the desired effect in its performance. To overcome the issues in currently available prediction techniques, this paper introduces an efficient software failure prediction technique using hybrid machine learning algorithms. First part of the work performs feature selection with an improved fitness function by utilizing genetic algorithm (GA) to optimize the features in the data set. After selecting the better features, Decision Tree algorithm is used as a classification technique for processing that features. The work compares the GA-DT based hybrid model with the currently available machine learning model such as RCSOLDA-RIR and WPA-PSO for the prediction of software failure. The outcome of the experimental analysis shows that the proposed model achieves better accuracy than the currently available model.

Topics & Concepts