Correlation-based modified long short-term memory network approach for software defect prediction

Developing software applications has become more perplexing nowadays due to the huge usage of software applications. Under such circumstances, developing software without defects is a very challenging task. So, detecting defects in software modules is necessary for the developers to allocate appropriate sources for the project. Knowing the defects in advance increases the software quality at a low cost. This article aims to develop a correlation-based neural network model for identifying defects in software projects. A novel correlation-based modified long short-term memory neural network (CM-LSTM) is proposed to estimate the software defects in software modules with modeled data. Based on the positive correlation between the features and the target variable, target variables have been changed. The prepared data is fed to the LSTM model to overcome the imbalance issue in the software defect prediction data. The adequacy of the proposed method is tested with a JM1 software defect prediction dataset with various performance parameters. It is observed that the proposed correlation-based modified LSTM technique is effective in detecting defects in software projects. The proposed technique employs correlation-based feature selection for long-short term memory neural networks to identify defects in software projects, and it is found to be more efficient than other existing approaches such as correlation-based LSTM, K-nearest neighbor, Stochastic gradient descent, Random forest, Gaussian Naive Bays, Logistic regression, Decision trees, Linear discriminant analysis, Multi-layer perceptron.