Application of STRIM to a Real-world Dataset and Proposal of Expanded STRIM
Yuichi Kato, Tetsuro Saeki · Transactions of the Institute of Systems Control and Information Engineers · 2023
We have previously proposed a statistical test rule induction method (STRIM), which induces the causality by if-then rules hiding in the dataset called the decision table in the field of the Rough Sets and confirmed its validity in a simulation model. However, the task of studying its validity and usefulness in a real-world dataset (RWD) was left to future research. Generally, the result of rule induction from an RWD cannot be directly ascertained. Therefore, after the previous STRIM was applied to an RWD, the induced rules were applied to the classification problem, and the result of classification was recognized as the validity of the rule induction method because the result was directly affected by the induced rules. Here, the classification result by Random Forest (RF) was used for an index of validity of that by the previous STRIM.