Towards developing an effective approach to component based software cost estimation

Abstract

newline quotSoftware project planning with estimation is a major and important activity carried out before developing any software. The development of software is broadly based on the project planning as the estimation of software plays a very vital role in successful completion of any software project with optimum time. newlineEstimating the cost and time of software in all aspect is a very tedious task. The same applies for the development and staffing required for software development. A person with the requisite skill, experience, and good thinking can make and take more precise decisions that will help in controlling and planning software risks, which are admirably correct and valid. The estimation of software cost and efforts are depended upon different parameters such as the size of the software, efforts taken by a person per month and the number of people working during the duration of the project. newlineThe research work focus on improving the accuracy for evaluating and comparing the efforts and software cost. The research works are mainly divided into two parts. The first part involves the development of a Hybrid model, which is the combination of the Cost Constructive Model-II, Artificial Neural Network, and Principal Component Analysis. In the model, Module-I determines the size, efforts, duration, people, and cost by using seventeen cost drives and five scale factors inputs of plain COCOMO-II model. The Module-II uses ANN to determine the efforts, duration, people, and cost using the same input as applied to Module-I through 42 input values and sample dataset. The Sample dataset is the brain of ANN, which consists of 10 project data, and each project includes 42 values. The weighted sum of 42 inputs combined with the threshold values is trained by using supervised learning algorithm called Delta Rule Learning. It uses trained data for calculating effort, time, people, and cost. Module-III is a combination of Module-I and Module-II with PCA.quot newline