Classification of Software Failure Incidents Using SVM

Abstract

Software failure in an operational environment can put the performance and quality of service at risk. This research is particularly on software failure incidents. The study involves the basic software engineering process: Classification of software failure incidents through machine learning techniques. The active learning approach is used, which is applied to label only those data which is most in-formative to build models. From all the samples, the sample with higher entropy (randomness) is chosen for labeling. Given a set of labeled observations, we used a classifier that decides the target class label, either “failure” or “no failure”. As a classifier mechanism, Support Vector Machine (SVM) is used to classify the data.

      Keywords:  incident, machine learning, active learning, SVM