Abstract:
Background: The necessity of computer grew rapidly. Consequently, the use of software is enormous and complex. Software companies are building increasingly complex systems. At the same time, the market needs them to complete their project in less time while customers also require high quality. Software companies have various measurement methods some of them are customer feedback, after it delivered to customers and software testing, before the software delivered to customers or stakeholder. Objective: The goal of this study is to apply combined machine learning methods for predicting the status of software faults from the NASA MDP data set. Especially, the researcher gives emphasize to build best prediction model, finally the researchers develop and evaluate the best model by building a prototype system. Methodology: In this Study, the methodology was considering both single and combined machine learning approach. This research work has a total of 498, 10885, 2109, 522 and 458 records and 22 attributes in order to predict software faults. Noise removal and handling of missing values was done to prepare the dataset for experiments. For the purpose of building the model machine learning algorithms such as decision tree, support vector machine, artificial neural network and combination of single classifiers using vote method is used. To evaluate the model accuracy 10-fold cross validation is used and Accuracy, precision and Recall was used to evaluate the performance of the developed model. Finding: The evaluation of best performing methods is compared according to accuracy, sensitivity, specificity and execution time to build the model. Based on performance evaluator the best algorithms are found to be combination of J48 with SMO classifiers were SMO is followed by J48 classifier. Contribution: information extraction from the source code should be done before the researcher an attempt to test or maintain the software is done; then the developed model will be applied for determining the complexity of the software product.