Predicting rare failure events using classification trees on large scale manufacturing data with complex interactions

Manufacturers collect large amounts of information during production. This data can be hard to interpret because of the presence of many unknown interactions. The goal of this study is to demonstrate how machine learning methods can uncover important relationships between these parameters. This paper describes several techniques to predict rare failure events. Single parameter methods are used to explore performance over time. Tree based classification methods are used to identify deep interactions that cannot be identified by linear methods. Tree methods are shown to identify non-linear relationships.