Classification of software defects is an important task to know the type of defects. It helps to prioritize the defects, to understand the cause of defects for improving the process of software defect management system by taking the appropriate action. In this paper, we evaluate the performance of naïve Bayes, support vector machine, k nearest neighbor, random forest, and decision tree machine learning algorithm to classify the software defect based on orthogonal defect classification by selecting the relevant features using chi-square score. Standard metrics accuracy, precision, and recall are calculated separately for Cassandra, HBase, and MongoDB datasets. The proposed method improves the existing approach in terms of accuracy by 5%, 20%, 6%, 27%, and 26% for activity, defect impact, target, type, and qualifier respectively, and shows the enhanced performance.