The Correlation Aspects of Software Development Actual Effort and the Effected Factors

Abstract

estimating the effort of software development is the focus of attention of software engineers, as it allows developers to determine the resources needed for the software project from the beginning until the final delivery. In fact, the actual effort of the software development depends on a set of basic factors that affect it either positively or negatively. These factors vary depending on the type and nature of the software. A few studies have been conducted in order to consider these factors and their correlation to the actual effort. Most of these studies have used methods such as a survey or a questionnaire. However, these methods lack of inaccurate measurement due to the variation of participated experiences and skills. Today, the world have emerged artificial intelligence technologies in most of fields to obtain higher accurate and fast results. This research uses feature selection intelligent algorithms such as Generic Univariate Select, Mutual Information, and RelifF algorithms to determine the correlation of software actual effort with the effected factors through XGBoost machine learning library. The NASA data set and the common effected factors of COCOMO model has been selected to determine the factors with the strongest correlation and influence on the actual effort. Through the results of this research, it has been found that complexity factor has recorded the largest correlation coefficient followed by the size of the software. The relationship between factors has been noticed and this contributes to the importance of the effected factors consideration and perfect software management.