The use of a genetic algorithm for clustering the weighing station performance in transportation – A case study

Abstract: In this paper, a genetic algorithm (GA) is developed to solve a clustering problem for evaluating and ranking the weighing stations according to their performances. In hierarchical steps of clustering, observations with the least similarities should be merged and some of them will be lost. To improve this defect, the main concept behind the proposed algorithm is to avoid losing data in the hierarchical process of clustering, so all of the observations are randomly assigned into a predefined number of clusters by GA procedures. In this model, we consider the performance factors related to the weighing operation, such as the traffic volume of trucks, detected overloading, type of portable or fixed scales, and rate of acceding detections compared to the same duration in the previous year. The required data of 126 weighing stations are collected during two 6-month periods. Different dimensions of the collected data are standardized to uniform dimensions. The main performance of a clustering method considered as the fitness value in a genetic algorithm (GA) is to maximize the sum of deviation squares from the mean of within groups. It guaranties that the clusters have most similarities within groups and least similarities in among groups. Four different techniques of the mathematical clustering are compared with the result of the proposed GA by using the MATLAB software. The related results show that the clustering of weighing stations is more likely to other methods. [Copyright &y& Elsevier]