Classifying Composition of Software Development Team Using Machine Learning Techniques

Abstract

Software development projects still reportedly have high failure rates. The ineffective composition of the software team has been recognized as the main aspect of the failure of the software project. In this study, a classification model of the composition of an effective software development team was developed. The model developed consists of three predictor variables: personality, role, and gender. Outcome variables to determine team effectiveness are seen in the quality of the team. To measure the quality of the team, two metrics were used: team development level assessment and team dysfunction assessment. The techniques used for classification are logistic regression and decision trees. The experimental results show that the best method is produced by a decision tree with the highest accuracy value of 70%. Therefore, the results conclude that the use of the decision tree method can determine an effective team as software development team.