A Computational Intelligence Strategy for Software Complexity Prediction

The automated prediction of software module complexity using quantitative measures is a desirable goal in the area of software engineering. A computational intelligence based strategy, stochastic feature selection, is investigated as a classification system to determine the subset of software measures that yields the greatest predictive power for module complexity. This strategy stochastically examines subsets of software measures for predictive power. Its effectiveness is measured against a conventional artificial neural network benchmark.