Novel Visualization XAI Techniques for Assisting Schedulers’ Comprehension and Interpretation of ACO Applications in Job Scheduling

Bio-inspired algorithms, such as ant colony optimization (ACO) and others, have been widely applied to job scheduling. However, ACO applications were viewed as black boxes that were difficult to understand, trust, and accept. To address this issue, this study proposes several novel visualization XAI techniques, including color-encoded disjunctive subgraphs, color-coded pheromone distribution maps, several variations of dynamic transition and contribution diagrams, and the application of contrastive gradient saliency maps to the dynamic line chart for ACO to pinpoint the convergence point of the evolution process to avoid further evolution rounds. The proposed methodology has been applied to a flexible job shop scheduling problem. Thirty-three experts were also asked to evaluate and compare the performances of various visualization XAI techniques using analytic hierarchy process (AHP). Experimental results showed that the color-encoded disjunctive subgraph reduced the execution time by 98% compared to the traditional disjunctive graph.