A Fuzzy-Based XGBoost Approach for Classification and Prioritization of Cost Overhead Factors in Agile Software Development

Abstract

As a result of these concerns with regard to cost overhead reduction, this paper examines the problematics of exact classification and prioritizing in Agile software development. Although using Agile iterative approach may add the levels of needed complexity in making smart decisions to control cost and resources. The work proposes an approach integrating fuzzy logic with XGboost to address these difficulties. Thus, while XGBoost bases categorisation and prioritisation of cost overhead variables on its gradient-boosting method, proven to reliably classify and rank factors, fuzzy logic is employed to respond to uncertain conditions characteristic of Agile environments. Compared to other approaches, the suggested hybrid model shows significant improvement on both counts. Greater accuracy in categorisation ensures more uniformity in the identification of cost overhead elements, which proves useful for decisions in Agile projects. Proper prioritisation also enables planers in the team to reduce time wastage and allocate available resources in the best ways possible. The criteria which serve to evaluate paves for the strength of the model. As revealed with better accuracy in training, handling complex data, the hybrid model does have higher accuracy than Decision Tree, Random Forest and SVC. That is why the optimisation of feature selection in XGBoost and the ability to handle uncertainty in fuzzy logic increase the accuracy of the latter. Through recall, the process is much boosted and as such ensures all elements are noted. Amongst the measures, the pure handicapped loss less exact assignment of the F1 score assesses that carries a proposing of both recollect and precision that establishes the general organization of the model through studying the Agile cost overhead control method.