Research On The Multiple Classifier Integration Based On Diversity Measure

Since multiple classifier systems can to some extent improve the performance of classification, the technique has been widely used in various fields of pattern recognition. The recent research on multiple is no longer limited to proposing or enhancing the integration methods. More and more attention is paid for the relationship among base classifiers, especially the diversity in classifier ensembles. Researchers want to fmd a proper diversity measure, whose ability in predicting the performance of classifer integration can be used to assist the construction of multiple classifier systems.This paper studies these two main problems in multiple classifier systems: integration methods and diversity measures. The main contributions of the work are as follows:Firstly, an algorithm of ensemble Selective based on Diversity Measure is proposed. The algorithm is mainly added choice of classifier in the traditional multi-classifier integration system, in other words, before multiple classifiers integrate, we first choose the new classifiers set. The experiments on UCI database prove that, using the algorithm, though the number of base classifiers in the classifier ensemble is reduced, the error rate can be maintained or even decreased.Secondly, Considered about the accuracy of individual classifier and diversity among individual classifiers, this paper proposes an ensemble algorithm based on diversity measure feature selection DMFS, which combines the feature selection technique, Relief, with diversity measure to find the optimal ensemble. By applying the DMFS on UCI data sets, the result demonstrates the DMFS algorithm achieve high speed, and its accuracy are both higher than Bagging and Boosting algorithm.Thirdly, the author considers content-based image retrieval as example to describe the application of DMFS, presents a new image retrieval method based on independent key-block. At the same time, apply the presented algorithm of diversity measure feature selective into the image classification and get well effect. In the end, this paper concludes by summarizing the research and indicating its future work.