Performance Analysis and Enhancement of Software Quality Metrics using Decision Tree based Feature Extraction

Software managers routinely face the problem with software projects that contain error or inconsistencies which increases the budget, time limit and reduces the customer satisfaction. By applying the data mining technique for software metrics dataset as a quality prediction model, helps manager to tackle the above problems in an efficient way and improve the quality. In this paper, Decision tree classifier, a data mining technique is used as the software quality prediction model and applied on the reduced dataset for evaluating the overall performance by measuring the Accuracy, Mean Absolute Error(MAE), Root Mean Square Error(RMSE), Precision and Recall. The observations show that the overall accuracy is increased when applied with the reduced dataset than the original data set. Hence we can term that the reduced dataset play a vital role in the improving the software quality by increasing the classifier performance and hence can be said that the attributes in reduced dataset are enough as good predictors for future classification. The dataset used here is the KC2 NASA data set. To evaluate the performance of decision tree classifier, measures such as Accuracy, MAE, RMSE, Precision and Recall are used.