Scheduling Feature Selection for Data-driven Job Shop Scheduling System Using Improved Firefly Algorithm Optimization

Abstract

This study proposes an improved version of the Firefly Algorithm (FA) for scheduling feature selection optimization in data-driven job shop scheduling system. The proposed FA variant employs a dynamic step strategy with the elite individual dipartition in firefly population to improve the optimization ability of FA. Selecting essential schedule feature data from production system attributes data based on various production requirements to construct data-driven job shop scheduling system is a critical issue because of the existence of irrelevant and redundant attributes in job shop production system, by selecting the important attributes as schedule features, robust performance can be expected in data-driven job shop scheduling system. A wrapped scheduling feature selection approach based on the proposed FA variant and extreme learning machine (ELM) is presented for ELM-based job shop scheduling system. The feasibility and effectiveness of the proposed scheduling feature selection approach have been verified via a practical job shop scheduling case.