Abstract:

For the multiple classifier integration in the pattern recognition, a decision forest rather than a decision tree was built to realize the submodel integration by mining the relevance  in the predictive attributes in the test sample and giving the distinct classification rule to each sample based on the conditional independence analysis of the training set. The structure and the number of the decision trees can be defined adaptively during the learning process. Experiments on UCI learning data sets proved the feasibility and effectiveness of the proposed method.

Key words: artificial intelligence, pattern recognition, decision forest, conditional independence assumption, data mining model