Analyzing Correlation of the Relationship Between Technical Complexity Factors and Environmental Complexity Factors for Software Development Effort Estimation

Abstract

In this paper, a new method called Correlation-based Feature Selection in Correction Factors is proposed. The method is based on the feature selection method used in software development effort estimation to reduce redundant correction factors. In this paper, the impact of correlation-based feature selection on the method’s estimation accuracy is investigated. Multiple linear regression was used as the basic technique for the correction factors preprocessed by the feature selection method. The results were evaluated using six unbiased accuracy measures through the 5-fold cross-validation over the historical dataset. The proposed method leads to a significant improvement in estimation accuracy by simplifying the evaluation of correction factor values in the use case points method, thus increasing the usefulness of the proposed method in practice.