A New Evidential K-Nearest Neighbors Data Classification Method
Yan Zhang · Fire Control and Command Control · 2013
The K-Nearest Neighbor(K-NN)rule has been widely used in the pattern recognition field.In order to effectively deal with the uncertain information and to improve the accuracy of classification,a new evidential K-Nearest Neighbors(NEK-NN)data classification method is proposed.Several training subsets are resampled from the whole training set.In each subset,the basic belief assignments(bba's)are determined using the distance between the object and its K Nearest Neighbors,and then the K bba's are discounted according to the number of the K Nearest Neighbors in each class.Finally the discounted bba's are combined using DS rule,and the mean of these combination results in each training subset is used for the classification of the object.Several experiments are given to test effectiveness of NEK-NN with respect to some other methods.The results indicate that NEK-NN can effectively improve the classification accuracy.