Predicting Software Effort Estimation Using Machine Learning Techniques

In software engineering, estimation plays a vital r ole in software development. Thus, affecting its cost and required effort and consequently influencing the overall success of sof tware development. The error margin in Expert-Based, Anal ogy-Based and algorithmic based methods including: COCOMO, Fu nction Point Analysis and Use-Case-Points is quite signifi cant, which exposes software projects to the danger of delays a nd running over-budget. To obtain better estimation, we propos e an alternative method through performing data mining o n historical data. This paper suggests performing this predictio n using three machine learning techniques that were applied to a preprocessed COCOMO NASA benchmark data which covered 93 project s: Naive Bayes, Logistic Regression and Random Forests. The generated models were tested using five folds cross -validation and were evaluated using Classification Accuracy, Preci sion, Recall, and AUC. The estimation results were then compared to COCOMO estimation. All the applied techniques were successful in achieving better results than the compared COCOM O model. However, the best performance was obtained using bo th Naive Bayes and Random Forests. Despite the fact that Nai ve Bayes outperformed both of the other two techniques in it s ROC curve and Recall score, Random Forests has a better Confu sion Matrix and scored better in both Classification Accuracy, and Precision measures. The results of this work confirm the vali mining in general and the applied technique in part software estimation.