Development of analogy-based estimation method for software development cost estimation in government agencies

Cost estimation for software development in Government Agencies is still considered as a challenge. Owner Estimate Cost (OEC) should be estimated based on the specifications at the early stage of the procurement. There is no standard used by the government in formulating technical specifications. This affects the OEC value that tends to be highly subjective. Therefore, it is important to develop an estimation method that is able to represent software complexity at the early stage of procurement. The approach that can be considered in the early stages is the Use Case approach. UCP is a method to calculating effort algorithmically using use case complexity. However, UCP has a limitation, which only provides a level of fixed complexity and cannot handle uncertain conditions. In the other hand, Some previous research result shows that the analogy method has better performance than the algorithmic method. In this study, we developed the effort and cost estimation methods based on analogy by building a new dataset using the use case complexity parameters. Datasets are collected from 100 historical software projects data that have been built using the UCP method approach. Cost components for the project follow the current procurement regulations in Indonesia. Dataset evaluation for effort estimation using the proposed analogy method shows the best results of MMRE of 0.36 and PRED(0.25) of 0.57. MMRE shows the average difference of actual effort and estimated effort, whereas PRED (0.25) shows prediction level with error value smaller or equal to 25%. Furthermore, the estimated cost of the three software projects resulted in an average percentage deviation of 7.37%.