Detection Defects in Software Testing Using Machine Learning Techniques

Abstract

Machine Learning is a branch of Artificial Intelligence that focuses on the development of systems that learn from their data. Machine learning has the capacity of collecting raw data from a repository and converting it into a form that can be used for calculation and prediction of software bugs. It is usually preferable to identify a software problem as soon as possible in order to minimise the amount of time and money spent on fixing it. Technique for determining characteristics It is necessary to apply the wrapper and filter technique in order to discover the most optimum software metrics. The primary goal of this article is to identify the most effective model for predicting software bugs. The goal of certain machine learning applications is to learn characteristics of data sets in situations when the right solutions are not previously known by human users. Because there is no trustworthy test oracle for machine learning software, it is difficult to test such software.

There are several machine learning techniques used in this study including linear regression and random forest. Other techniques include neural networks, support vector machines, decision trees, and decision stumps, and a comparative analysis has been conducted using performance parameters such as correlation, R-squared, mean square error, and accuracy for software modules such as ant, ivy, tomcat, berek, camel, lucene, poi, synapse, and velocity. When compared to other machine learning models, the support vector machine outperforms them all. Traditionally, the emphasis of software defect prediction has been on the creation of static code metrics, which, when fed into machine learning classifiers, aid in the prediction of the fault probability of a code. The location of errors in code bases has been predicted using machine learning methods such as Deep Learning methodology and Ensembling, Data Mining, Clustering, and Classification. However, researchers have not yet agreed on the optimal prediction model to use. The findings demonstrate that consistency in high accuracy prediction may be obtained via the use of Ensembling methods.