Improving the Performance of Heuristic Algorithms Based on Exploratory Data Analysis

This work tries to promote the application of empirical techniques of analysis within computer science in order to construct models that explain the performance of heuristic algorithms for NP-hard problems. In this paper, we show the application of an experimental approach that combines exploratory data analysis and causal inference with the objective to explain the algorithmic optimization process. As a case study we present an analysis of some heuristics algorithms for the Bin Packing Problem (BPP). The knowledge gained about the structure of BPP, the heuristics algorithms and the relationships between the characteristics that define them, can be used to: a) classifying instances of BPP by degree of difficulty, b) explain the performance of the algorithms to different instances c) predict the performance of the algorithms for a new BPP instance, and d) develop new strategies of solution.