Neural network approach for software cost estimation

Software engineering measurement and analysis specifically, cost estimation initiatives have been in the center of attention for many firms. In this paper, author explores the use of the expert judgment and machine learning techniques using neural network as well as referencing COCOMO II approach to predict the cost of software. Some primary work in the use of neural network in estimating software cost by [WITTIG & FINNIE] and Karunanithi (1992) produced very accurate results, but the major setback in their work was due to the fact that the accuracy of the result relied heavily on the size of the training set (1997). Understanding the adversity in applying neural networks, the author proposes a dynamic neural network that initially uses COCOMO II. The proposed network improves its estimation as the number of data set increases with input from expert judgment that affects the learning procedure.