Time Effort Prediction Of Agile Software Development Using Machine Learning Techniques

Abstract

To run a software development project, an effective and efficient project management mechanism is needed to coordinate the activities carried out. The agile method was developed because there are several weaknesses in the classic method that can interfere with the course of the software development process according to user desires.  However, in applying agile methods, time effort estimation cannot be done properly. This can cause project managers to have difficulty preparing resources in software development in scrum projects. For this reason, this research aims to predict the time effort of agile software development using Machine Learning techniques, namely the Decision Tree, Random Forest, Gradient Boosting, and AdaBoost algorithms, as well as the use of feature selection in the form of RRelieff and Principal Component Analysis (PCA) to improve prediction accuracy. The best-performing algorithm uses Gradient Boosting k-fold validation with PCA with an MSE value of 2.895, RMSE 1.701, MAE 0.898, and R2 0.951.