Multistage software estimation

This paper proposes a COCOMO-based multistage software estimation approach. In the first stage, Monte-Carlo simulation is used to identify the key cost drivers in COCOMO II model, which contribute most to the variance of the software effort prediction. In the second stage, six selected cost drivers, as fuzzy inputs, are fed into three fuzzy expert systems to get the effort multiplier estimation. An empirical case study is used to evaluate the simulation results. Numerical data analyses show that the proposed multistage fuzzy logic model is quite encouraging compared to original COCOMO II approach.