In the pattern recognition, it is very effective to unite the output of many classifiers when we deal with complex classification questions. This kind of union is called classifier ensemble, and the fuzzy integrals is such a kind of methods. The classifier ensemble based on Choquet integrals can effectively overcome the uncertainty of single classifier, so that it can obviously enhance the precision and the compatibility of the classification model. In recent years, the method of classifier ensemble based on Choquet integrals has been widely used in the research of homogeneous classifier ensemble, and this has obtained very good results. But it is not common that this method is being used in heterogeneous classifier ensemble. The main purpose of this article is to study the model of neterogeneous classifier ensemble based on Choquet integrals. At first we introduce the elementary knowledge of fuzzy integrals and the classifier ensemble. Then we propose the concept of neterogeneous classifier ensemble. Furthermore we propose the model of neterogeneous classifier ensemble based on Choquet integrals and give the concrete computational methods and steps. At last we test our model with eight data sets and analyse the experimental results with statistical methods. The main conclusions are: the ensemble models generally surpass the single classifier models; The ensemble models based on Choquet integrals surpass the simple average ensemble model; In the contrast of two ensemble models based on Choquet integrals, the improved method of determining the fuzzy density surpasses the method of using the correct rate as the fuzzy density. This result on the one hand confirms that the multi-classifiers ensemble models are generally better than the single classifier models. On the other hand this specifies that the Choquet integrals is an excellent classifier ensemble tool. Not only it is very useful in the homogeneous classifier ensemble, but also it has a good performance in the neterogeneous classifier ensemble.
Download Info
To download:
If you experience problems downloading a file, check if you have the
proper application to
view it first. Information about this may be contained
in the File-Format links below. In case of further problems read
the IDEAS help
page. Note that these files are not on the IDEAS
site. Please be patient as the files may be large.
Publisher Info
Paper provided by University Library of Munich, Germany in its series MPRA Paper with number
10414.