Development of a Rule Evaluation Support System with Adaptive Rule Evaluation Model Learning

点击翻译按钮获取中文摘要

In this study, I assumed the iterative situation about if-then rule evaluations by human experts in post-processing phase of data mining. From the results of my empirical studies, we can get the following lessens-of -learn: (1)the initial cost to build rule evaluation models can be reduced, because the method only needs small balanced subsets of labeled rules sampled by using the accuracy of trial rule evaluation models. (2) At the initial rule evaluation iteration, a human expert can use some functional groups of objective rule evaluation indices to measure rules. And, the system can also build rule evaluation models by using these groups of indices. (3) The transitions of criteria of human experts can be detected by measuring rule evaluation models consisting of if-then rules.