A visualization tool for heuristic algorithms analysis

Abstract

The performance of the algorithms is determined by two elements: efficiency and effectiveness. In order to improve these elements, statistical information and visualization are key features to analyze and understand the significant factors that affect the algorithm performance. However, the development of automated tools for this purpose is difficult. In this paper a visual diagnosis tool named VisTHAA, which provides researchers statistical and visual information about instances and algorithms, is proposed. Besides, VisTHAA allows researchers to introduce characterization measurements, make algorithm comparisons with a non-parametric test and visualize the search space ruggedness and the behavior of the algorithm. Due to the above, in this study we used VisTHAA as a tool for improving the efficiency of a reference algorithm in the literature. In the study case, the experimental results showed that through visual diagnosis it was possible to increment the algorithm's efficiency to 93% without a considerable loss of effectiveness. © 2013 Springer-Verlag.