Comparison of Data Mining Techniques in the Cloud for Software Engineering

Abstract

Mining software engineering data has recently become an important research topic to meet the goal of improving the software engineering processes, software productivity, and quality. On the other hand, mining software engineering data poses several challenges such as high computational cost, hardware limitations, and data management issues (i.e., the availability, reliability, and security of data). To address these problems, this chapter proposes the application of data mining techniques in cloud, the environment on software engineering data, due to cloud computing benefits such as increased computing speed, scalability, flexibility, availability, and cost efficiency. It compares the performances of five classification algorithms (decision forest, neural network, support vector machine, logistic regression, and Bayes point machine) in the cloud in terms of both accuracy and runtime efficiency. It presents experimental studies conducted on five different real-world software engineering data related to the various software engineering tasks, including software defect prediction, software quality evaluation, vulnerability analysis, issue lifetime estimation, and code readability prediction. Experimental results show that the cloud is a powerful platform to build data mining applications for software engineering.