A New Fuzzy Clustering Based Method to Increase the Accuracy of Software Development Effort Estimation

Project planning plays a significant role in software projects so that imprecise estimations often lead to the project faults or dramatic outcomes for the project team. In recent years, various methods have been proposed to estimate the software development effort accurately. Among all proposed methods the non algorithmic methods by using soft computing techniques have presented considerable results. Complexity and uncertain behavior of software projects are the main reasons for going toward the soft computing techniques. In this paper a hybrid system based on combining C-Means clustering, neural network and analogy method is proposed. Since, there are complicated and non linear relations among software project features, the proposed method can be useful to interpret such relations and to present more accurate estimations. The obtained results showed that fuzzy clustering could decrease the negative effect of irrelevant projects on accuracy of estimations. In addition, evaluation of proposed hybrid method showed the significant improvement of accuracy as compared to the neural network the analogy method and statistical methods.