Predicting Effort and Errors for Embedded Software Development Projects 2. Software Project Management and Issues 3. Software Development Processes and Selection of Data 3.1 Data Sets for Creating Models

Recently, growth in the information industry has caused a wide range of uses for information devices, and the associated need for more complex embedded software, that provides these devices with the latest performance and function enhancements (Hirayama (2004); Nakamoto et al. (1997)). Consequently, it is increasingly important for embedded software-development corporations to ascertain how to develop software efficiently, whilst guaranteeing delivery time and quality, and keeping development low costs (Boehm (1976); Tamaru (2004); Watanabe (2004)). Hence, companies and divisions involved in the development of such software are focusing on various types of improvement, particularly process improvement. Predicting effort requirements of new projects and guaranteeing quality of software are especially important, because the prediction relates directly to costs, while the quality reflects on the reliability of the corporation Komiyama (2003); N. (2004); Nakashima (2004); Ogasawara & Kojima (2003); Takagi (2003). In the field of embedded software, development techniques, management techniques, tools, testing techniques, reuse techniques, real-time operating systems and so on, have already been studied. However, there is little research on the relationship between the scale of the development and the number of errors, based of data accumulated frompast projects. As a result, previouslywe described the prediction of the total scale usingmultiple regression analysis (Iwata et al. (2006b); Nakashima et al. (2006)) and collaborative filtering (Iwata et al. (2006a)). In this Chapter we therefore, propose a method for creating effort and errors prediction model using an Artificial Neural Network (ANN) for complementing missing values (Iwata et al. (2006a)). The proposed method calculates the amount of effort and the number of errors by the following 3 steps. The first step, the similarity between the complementary project data, which include missing values, and the complete project data is calculated. Next, applies collaborative filtering using the method Tsunoda et al. (Tsunoda et al. (2005)) to complement missing values in the data and thus produce sufficient amount of data. In the final step, the prediction target project effort (or errors) is calculated Applying an Artificial Neural Network to Predicting Effort and Errors for Embedded Software Development Projects 13