A Hybrid Associative Classification Model for Software Development Effort Estimation

A mathematical model that makes use of data mining and soft computing techniques is proposed to estimate the software development effort. The proposed model works as follows: The parameters that have impact on the development effort are divided into groups based on the distribution of their values in the available dataset. The linguistic terms are identified for the divided groups using fuzzy functions, and the parameters are fuzzified. The fuzzified parameters then adopt associative classification for generating association rules. The association rules depict the parameters influencing the software development effort. As the number of parameters that influence the effort is more, a large number of rules get generated and can reduce the complexity, the generated rules are filtered with respect to the metrics, support and confidence, which measures the strength of the rule. Genetic algorithm is then employed for selecting set of rules with high quality to improve the accuracy of the model. The datasets such as Nasa93, Cocomo81, Desharnais, Maxwell, and Finnish-v2 are used for evaluating the proposed model, and various evaluation metrics such as Mean Magnitude of Relative Error, Mean Absolute Residuals, Shepperd and MacDonell’s Standardized Accuracy, Enhanced Standardized Accuracy and Effect Size are adopted to substantiate the effectiveness of the proposed methods. The results infer that the accuracy of the model is influenced by the metrics support, confidence, and the number of association rules considered for effort prediction.