Fuzzy Cognitive Mapping Analysis to Recommend Machine Learning-Based Effort Estimation Technique for Web Applications

Effort estimation is a fairly researched field in the area of software engineering. Algorithmic and non-algorithmic methods are the two popular ways of estimating software development efforts. Various machine learning techniques are also being used to determine project efforts based on the historical project-related dataset. These techniques consume an array of project characteristics to estimate the project cost. The selection of the right technique to correctly determine the project cost is a significant challenge that the software industry is facing. This paper presents a fuzzy cognitive mapping (FCM) approach to recommend the best machine learning-based software estimation technique for Web applications. FCM shows synergistic interactions between system variables, and this property is used in the context of Web application estimation for suggesting an estimation technique based on the Web project configuration. To counter the ambiguity in defining abstract relationships between system variables, this article also proposes to incorporate fuzzy numbers. The current analysis involves using five different estimation techniques on 125 student project records. The mean square error (MSE) was taken as a performance metric to declare the supremacy of one estimation technique over others. The experimental results show that the selection of an effort estimation technique should not ignore the presence of project characteristics in the input vector. The achievement of this work is that the proposed technique is capable of recommending the suitable most Web estimation model based on project credentials for a specific Web project; it refrains from suggesting an estimation model optimum for the most project configurations. The FCM approach on software estimation technique recommendation results in a probability of success equals to 70%.