| 摘要 |
A method, system, and computer program product for building a classifier based on the concepts similar to a Bayesian classifier, but using fuzzy theory provides better understanding and an improved approach to modeling the data. The concept of prior information about the class is eliminated, and focuses on the likelihood function (obtained from training data). This is modeled as fuzzy sets. The process of classification uses the possibility-probability transformation. A method of performing data analysis processing includes constructing a fuzzy likelihood by modeling a likelihood of data given a training class using fuzzy sets and transforming the fuzzy likelihood into a probability distribution using a possibility-probability transformation. |
