Evaluation of Machine Learning Algorithms on Software Fault Estimation (SFE)

Software Fault Estimation (SFE) approaches are being utilized to develop error free and cost friendly software. In this regard, machine learning algorithms are utilized to efficiently estimate the faults in the software. In this study, we implement eight machine learning based algorithms such as Long Short Term Memory networks (LSTM), K-Nearest Neighbors (KNN), Recurrent Neural networks (RNN), Graph Neural Network (GNN), Multi Layer Perception (MLP), Naïve Bayes, Random Forest, and Entropy using logistic regression model. All the algorithms are applied to the promise dataset for performance evaluation based on performance benchmarks such as recall, accuracy, and precision. MLP and LSTM algorithms showed promising results as compared to the other algorithms in terms of precision, accuracy, and recall. LSTM showed 0.83,0.85, and 0.8% precision, accuracy,and recall, respectively whereas RNN showed 0.79%, 0.60, and 0.79% recall, precision, and accuracy, respectively.