Fusion of Rule-Based and Sample-Based Classifiers - Probabilistic Approach

摘要

The present paper is devoted to the pattern recognition methods for combining heterogeneous sets of learning data: set of training examples and the set of expert rules with unprecisely formulated weights understood as conditional probabilities. Adopting the probabilistic model two concepts of recognition learning are proposed. In the first approach two classifiers trained on homogeneous data set are generated and next their decisions are combined using local weighted voting combination rule. In the second method however, one set of data is transformed into the second one and next only one classifier trained on homogeneous set of data is used. Presented algorithms were practically applied to the computer-aided diagnosis of acute renal failure in children and results of their classification accuracy are given.