Adaptive Evolutionary Fuzzy Inference System for Accurate Software Project Planning and Cost Prediction

Abstract

This research suggests a hybrid computational model of most appropriate software project planning cost estimation through the combination of fuzzy logic and genetic algorithm optimisation with machine learning models. The study uses three systematic datasets used to show different software development settings to test the predictive ability of the suggested method. Gene elimination and genetic optimisation, recursive feature elimination, data preprocessing and feature extraction are used to improve the relevance of features and dimensionality reduction. The fuzzy inference mechanisms are used to model the uncertainty of the project characteristics like complexity, duration and team experience and the genetic algorithm is used to optimise the membership functions and rule weights so that the error in estimation is reduced to a minimum. Experimental evidence confirms that the hybrid model is always better in lowering the values of MAE and RMSE and increasing the values of R 2. The results are that uncertainty-conscious reasoning with evolutionary optimisation leads to better prediction accuracy and robustness, which makes the model applicable in the cost estimation and decision support of software projects that are expected to be reliable even in the dynamic development settings.