Enhancing the multiple regression-based software cost estimation modeling process by adopting bootstrapping method

Abstract

As the system functionalities implemented by software and the size of software increase, it has become important in project management to estimate software development cost accurately. Hundreds of models have been developed for accurate software cost estimation over the last few decades. Among these models, multiple regression-based models are widely used in practice because these can give relatively accurate results. However, these regression-based models can be easily affected by outliers, especially when the sample size is small. There exist no methods to develop software cost estimation models considering small sample size problem. Therefore, it is not easy to develop a highly predictive cost estimation model by using these in the case of a small sample size. In addition, there are only a few researches presented for the modeling methodology on developing software cost estimation model. Existing modeling methodologies do not present the process framework to develop model in a systematic and detail way. In this thesis, we propose a modified bootstrapping technique to develop cost estimation models when the sample size is small, which can increase the sample size and improve the estimation accuracy. Then, we propose a generic software cost estimation modeling process (SCEMP) which combines the existing modeling methodologies with this proposed bootstrapping method. Validation of the SCEMP was carried out by using the data collected from the development of weapon systems. Experiment results showed that the estimation accuracy was increased by application of the proposed modeling process. Additionally, we performed an empirical study to identify an appropriate sample size when we can apply the bootstrapping method. It is expected that SCEMP can help the model developer to build accurate software cost estimation models in a systematic way and contribute to the software project management, such as project planning, scheduling, risk analysis, decision making, and...