Background and Objectives: Nowadays, data mining is one of the most significant issues. One field of data mining is a mixture of computer science and statistics which is considerably limited due to increase in digital data and growth of computational power of computers. One of the domains of data mining is the software cost estimation category.
Methods: In this article, classifying techniques of learning algorithm of machine and COCOMO model as the most common estimation model of software costs are presented. Then, the analysis method of principal component approach is presented.
Results: This article presents a suitable method to improve the performance of the software cost estimation. Moreover, the basic data set is decreased and is turned into a new collection by using this method. Among the features, the best are extracted. The algorithms of several classifications are assessed by applying this method. Finally, the evidence for accuracy of our claims in terms of increase in estimation accuracy of software costs is presented.
Conclusion:. The results proved that the suggested method could have significant influence on models of decision tree, naïve Bayes and nearest neighborhood by decreasing dimension of input data and turning it into data.