Software Development Effort Estimation Using a Neuro Network Approximation Approach

The article describes the main effort estimation models for software development. It is spoken in detail about the most widely used software effort estimation model, the Constructive Cost Model (COCOMO). An approach to improve the accuracy of estimates of COCOMO model based on neural network approximation is proposed. It deals with the choice of a neural network with back-propagation errors as an approximator. Data are given about numerical results of neural network learning using COCOMO model parameters as input.