EXPERIENCES IN THE USE OF TECHNIQUES SOFTCOMPUTING IN THE EVALUATION OF SOFTWARE PROJECTS

Abstract

This paper presents application results of soft computing techniques in project´s evaluation. The authors analyze different project management´s guides regarding decision making techniques inside its. In this paper the authors combine traditional techniques for project´s evaluation with soft computing techniques. The authors propose an algorithm that combine the following techniques: traditional indicators from project management guides, fuzzy inference systems, machine learning algorithms and linguistic data summarization methods. In section number 3, the paper presents the comparisons of algorithms Anfis, Hyfis, FS.HGD, based on neural networks with algorithms GFS LS RS, GFS THRIFT based on evolutionary algorithms. In experiments, we applied cross validation techniques combined with Friendman test and Wilcoxon test Afterwards, the authors present an algorithm that build linguistic summaries to detect causes of project management failures. Finally, we present a comparison of the results and the conclusions.