Optimizing the Bayes Error in Classification Problem and Applying in Medicine

The efficiency of the classification problem by Bayesian method depends on the prior probability and the probability density function. This study proposes a method to find a prior probability based on the fuzzy cluster analysis problem, and estimate the probability density function by Copula method. Combining these two improvements, we can reduce classification error. Surveying on two different medical data sets about the number of elements, variables, and properties, the proposed method has shown the outstanding results in comparing to other popular ones. These specific numerical examples also show the potential in practical applications of this study, especially the medicine.