Segmentation Combination Method for Corrective Action Recommendation

The identification of issues in shoe assembly products may exhibit variations over time, necessitating real-time recommendations for corrective actions, which are of paramount importance. This paper introduces a novel approach that fuses two methods for segmentation-based corrective action recommendations. This approach is pursued because traditional Collaborative Filtering (CF) methods solely provide suggestions to managers based on trends of issue categories observed in shoes that share similar preferences with the target footwear. However, CF fails to account for potential dynamics in issue behaviors over time. In contrast, rule-based sequential recommendation techniques meticulously analyze the evolution of issue behaviors across time to unveil sequential rules in the structure of issue behavior during the preceding period and issue behavior during the current period. If the historical behavior of the targeted issue corresponds to the conditional segment of the rule, then the current-period issue behavior is categorized as the resultant segment of the rule. Sequential rules are derived from the sequence of corrective action groups for each issue category set, thereby furnishing valuable recommendations. Experimental results underscore the efficacy of this segmentation-based approach, yielding a recall value of 89.6%, precision reaching 88.9%, and an F1 score of 89.2%. These findings exemplify the superiority of this segmentation-based method over conventional CF and SR techniques.