A Rule Discovery Support System for Sequential Medical Data,-In the Case Study of a Chronic Hepatitis Dataset-
Miho Ohsaki · Medical Entomology and Zoology · 2002
It is needed for evidence-based medicine to support a medical expert in discovering clinically useful knowledge with data mining techniques. However, the real data on medical test results are severely intractable since they are sequential, large-scale, and ill-defined with many attributes and missing values. This paper discusses how pre-processing should be going and how a rule discovery support system should be developed and actually discovers medically interesting rules from the dataset on chronic hepatitis diagnosis. We have done the following preprocessing: unifying different names to the same entities, unifying different inspection cycle, discretizing time-series, and so on. Taking a general framework of time-series data mining based on pattern extraction and decision tree, we have discovered the rules consisted of the combination of medical test result patterns. The system has discovered medically interesting rules, and a medical expert has polished up the knowledge inspired by them through the iteration of rule discovery and evaluation.