Design failures often incur substantial cost overruns in the shipbuilding industry. Precautions against possible design failures facilitate on-time delivery and improved productivity. However, few studies have investigated the use of accumulated knowledge to prevent ship design failure. In addition, existing associative classification (AC) methods pay little attention to the rule consolidation process whereby discriminative association rules can be aggregated. In this study, we propose a new AC method that considers both support and confidence, while the number of matching features is taken into account not only to identify specific rules that capture useful associations, but also to enhance predictive performance by effectively aggregating relevant rules. We present an empirical case for the Korean shipbuilding industry by applying the proposed method to help reduce design failures by providing a designer with the most relevant revision history for a given design task so that unnecessary rectification can be avoided. The comparative results showed that the proposed method returns the best prediction accuracy among competing AC processes.

设计失误通常会导致造船业的大量成本超支。对可能的设计失败的预防措施有助于按时交货并提高了生产率。但是，很少有研究调查使用积累的知识来防止船舶设计失败。此外，现有的关联分类（AC）方法很少关注规则合并过程，从而可以区分出区别性的关联规则。在这项研究中，我们提出了一种新的AC方法，该方法同时考虑了支持和置信度，同时考虑了匹配特征的数量，不仅可以识别捕获有用关联的特定规则，还可以通过有效地汇总相关规则来增强预测性能。通过为设计者提供给定设计任务的最相关的修订历史记录，从而避免不必要的纠正，应用所提出的方法来帮助减少设计失败，从而为韩国造船业提供了一个经验案例。比较结果表明，所提出的方法在竞争性交流工艺中具有最佳的预测精度。