Paper Title
Estimating Software Cost using Multilayer Neural Network

Abstract
At Present, investing the cost in software projects are the most challenging task in the procedure of project management. Software cost estimation is used to evaluate the time and cost which are important to building software structure [1]. There are the different type of cost estimation methods, each method has their own pros and cons during the evaluation of development time and cost. The general question arising in our mind during the cost estimation is how to make them correct, as customers do not explicitly share their needs. No procedure is better or worse than others, in fact, their strengths and weaknesses are often admired for each other [3]. The accuracy of the assessment directly affects the success or failure of the project [4]. Accurate estimation not only saves time and resources when applications are updated or developed but also accelerates the update or development process [5]. This paper deals with different number of research papers, and it was concluded that researchers used different methods to estimate the software cost where calculated cost is not too near to the actual cost. So this paper focused on multilayer neural network to estimate the software cost with efficient result. The COCOMO 81 dataset has been used to train and test the network. The outcomes of the trial of trained neural networks are compared with that of COCOMO model. The determination of our research is to increase the accuracy of the estimation of the COCOMO model by introducing it to the Multilayer neural network [6]. From our experimental results, it was resolved that the proposed Multilayer Neural Network model diminish the Magnitude Relative Error (MRE) between actual cost and calculated cost. This paper also demonstrates the design and implementation of software cost estimation using multilayer neural network and its dream to extend this work using different dataset. Keywords - Software Cost Estimation; COCOMO Model; Multilayer feed-forward Neural Network; Activation Function; Magnitude Relative Error(MRE).