Analysis of Accuracy on Data Visualization Techniques for Multi-objective Algorithm Performance Based on Convergence and Diversity Towards the Pareto Frontier

Understanding the behavior of strategies that generate solutions for optimization problems with three or more objectives, such as evolutionary algorithms, is crucial for researchers. One practical approach to achieving this understanding is by the comparison of visualizations of the set of solutions that approximate the set of optimal solutions that address optimization problems with three or more objectives. Visualizations support the depiction of the approximate Pareto front shape, the establishment of relationships between objectives, the distribution of solutions, and the representation of convergence and diversity levels of algorithms. This chapter reviews the visual representations of algorithm performance with data of up to four dimensions, mainly analyzing convergence and diversity using Scatter Plot Matrix (ScPM), Heatmaps (HM), Parallel Coordinates Plots (PCP), and Radar Plots (RP) with state-of-the-art visualization techniques. As a result, this work contributes with identified shortcomings for a good visualization based on features such as accuracy, clearness, empowerment, and conciseness; also, it guides on how to use the aforementioned visual representations to reveal characteristics of the approximate Pareto sets and the distributions of the solutions.