不良品の因果関係に基づくグループ化による製品評価精度向上に関する研究

抄録

We proposed to establish measures by combining decision tree analysis and Neural Network for problems with small amounts of data. We used the decision tree analysis to extract the relationship of defective product data, and used the relationship to apply it to Neural Network. Compared with the method using the relationship of parameter, the method using the relationship between the classified units is considered to reduce the risk of noise generation in extracting. As a result, it was expected that high evaluation accuracy can be obtained when extraction is performed focusing on one defective group.