CASE-STUDIES IN HIGH-DIMENSIONAL CLASSIFICATION

We consider the application of several compute-intensive classificatio n techniques to two significant real-world applications: disk drive ma nufacturing quality control and the prediction of chronic problems in large-scale communication networks. These applications are characteriz ed by very high dimensions, with hundreds of features or tens of thous ands of cases. The results of several learning techniques are compared , including linear discriminants, nearest-neighbor methods, decision r ules, decision trees, and neural nets. Both applications described in this article are good candidates for rule-based solutions because huma ns currently resolve these problems, and explanations are critical to determining the causes of faults. While several learning techniques ac hieved competitive results, machine learning with decision rule induct ion was most effective for these applications. It is demonstrated that decision (production) rule induction is practical in high dimensions, providing strong results and insightful explanations.