Software Project Cost Estimation with Artificial Neural Networks

The aim of this thesis is to find out how can Artificial Neural Networks (ANN) – one of
the methods of Artificial Intelligence (AI) – be used in software projects cost estimation
and to give an example for one of the most appropriate models of ANN used for this

The estimation of effort for software projects, which includes both time estimation and
cost estimation, determines the project‟s destiny in a project management process.
Generally costs of software development are out of control because of the lack of
measuring and estimation methods. Although software project cost estimation is usualy
done by algorithmic methods such as COCOMO (Constructive Cost Model) and SLIM
(Software Life Cycle Management), recently researchers tend to use ANN to make more
accurate estimation. ANNs can be thought of as functions in the sense that they map a
set of inputs to outputs.

After a detailed information about AI and ANNs, we present the proposed ANN model
that will be used as estimator and give explanation for the web site that integrate both
the model and the user interface created to be used for the input of companies project

The results of this study show that the proposed ANN produce acceptable estimations.
If it will be trained and tested in order to be more precise, the results will be very close
to the actual project costs.

When this study was shared with some software companies, we saw that there is a
resistance in applying every new technology even among the software companies. But,
we believe that this study will be a starting point in applying AI and ANN methods to
real life applications, especially in software industry, in order to simplify the hard work
of project managers by reducing the time and effort in project cost estimation.