Visualization XAI Techniques for Facilitating Schedulers’ Comprehension and Interpretation of GA Applications in Job Scheduling
Toly Chen, Chi‐Wei Lin, Yu-Cheng Wang
Abstract
There are few applications of explainable artificial intelligence (XAI) in job scheduling. Various bionic computing algorithms, especially genetic algorithms (GA), have been applied to job scheduling. However, these GA applications were viewed as black boxes as they were difficult to understand, trust, and accept. Solving this problem is considered a viable way to improve the effectiveness of job scheduling further. To this end, this study first reviews existing visualization techniques and tools for explaining the applications of GAs in job scheduling, and then proposes several novel applications of existing visualization XAI techniques to enhance the effectiveness of explanations, including twin color-encoded chromosomes, saliency maps, and contrast gradient saliency maps. The proposed methodology has been applied to a real-world case of scheduling a flexible job shop with release time constraints using a GA.