A hybrid improved case-based reasoning approach and metaheuristics for cost estimation

Abstract

This study developed a cost predicting model based on a combination of Case Based Reasoning (CBR) and Genetic Algorithm (GA) to optimize the impact of attributes that forecast the installation cost for a new production line. This model will retrieve the old data to calculate the similarity against the problem to be solved, from which the prediction cost will be given. The goal of the model is to minimize forecast errors while finding the weights of the corresponding attributes. Besides GA, Particle swarm optimization (PSO) is considered in determining the impact of features to improve error rates. The results from GA and PSO will be compared to find the metaheuristics method that is accordant to the properties of the data set. Some analysis is also carried out to evaluate the sensitivity of specific attributes and parameter to the installation cost.