Application of a Multi-Objective Approach and Sequential Covering Algorithm to the Fatigue Segment Classification Problem

摘要: In this paper, a multi-objective rule discovery approach is introduced to address the problem related fatigue data editing. A set of rules are simplify editing process, with which users can easily predict an appropriate level damage for segments subject SAE1045 steel without need cycle-counting algorithm or stress analysis. called elitist non-dominated sorting in genetic (NSGA-II) modified and applied find high-level knowledge representation IF-THEN eight-dimensional search space. Three possible outcomes represented labels very low, low high chosen consequent. Two conflicting classification objectives, predictive accuracy comprehensibility, considered as optimisation criteria. To accelerate searching sequential covering advance training session, positively reduces size provides several simple rules. seven high-accuracy high-interpretability discovered, their uncomplicated implementation alternative task estimating value could be part process.