Rough set-based text mining from a large data repository of experts’ diagnoses for power systems

Abstract

Usually it is hard to classify the situation where uncertainty of randomness and fuzziness exists simultaneously. This paper presents a rough set approach applying fuzzy random variable and statistical t-test to text-mine a large data repository of experts’ diagnoses provided by a Japanese power company. The algorithms of rough set and statistical t-test are used to distinguish whether a subset can be classified in the object set or not. The expected-value-approach is also applied to calculate the fuzzy value with probability into a scalar value.