TOWARDS MEASURING EFFECTIVENESS IN DYNAMIC ENVIRONMENTS

Abstract

Many real-life scenarios can be modeled as Dynamic Optimization Problems (DOPs), which demand for find- ing optimal solution over time. From the viewpoint of metaheuristics methods, DOPs have been extensively addressed over the last two decades. One important issue in this context is how to assess the algorithm perfor- mance. Most of current proposals rely on single information from data, which limits the notion about the overall performance of the algorithm. So, in order to contribute to this issue, in this paper we propose a new performance
measure for algorithm assessment in evolutionary dynamic optimization. We derived our proposal from what we considered as effectiveness in dynamic environments. Different from other existing measures, our proposal involve not only the accuracy, but also the time (efficiency) of the algorithm. In order to illustrate its usefulness and relationship with other literature measures an experimental analysis was conducted. Results show that the proposed measure can be suitable employed in typical experimentation scenarios and offers new information about the algorithms performance.