NPKNN-SVM classifier based on Neyman-Pearson criterion

作者：Ye Qiaolin^*; Ye Ning; Hu Jie; Zhang Xunhua; Wu Bo

来源：Journal of Computational Information Systems, 2008, 4(5): 1909-1916.

• Conventional methods
• Feature space
• K nearest neighbours (k-NN)
• k-NN algorithm
• KNearest Neighbour (KNN)
• Neyman-Pearson criterion
• NPKNN-SVM algorithm
• Practical method
• Support vector
• SVM algorithm
• SVM classifiers
• SVM-KNN (KSVM) algorithm
• Testing samples
• Training sets

SVM-KNN (KSVM) classifier - A new method of improving the accuracy of SVM classifier is a better method of classification than conventional methods of SVM, which combines support Vector Machine (SVM) with KNearest neighbour (KNN), In the class phase, the algorithm computes the distance from the testing sample to the optimal hyperplane of SVM in feature space. If the distance is greater than the given threshold, the testing sample would be classified by SVM; otherwise, the KNN algorithm will be used. In fact, in most cases, the testing samples close to the optimal hyperplane which can be correctly classified by SVM are misclassified, so this algorithm may result in poor testing correctness. By KSVM algorithm, proposed in this paper is an improving NPKNN-SVM algorithm: The testing samples from whose distance to the optimal hyperplane is less than the given threshold are separated into two regions, then we present a practical method-N-P Process by Neyman-Pearson criterion so as to attain the larger number of testing samples which may be misclassified by KNneareast Neighbour (KNN) for KSVM, and support vectors are used as the training set in the test. The experiments show NPKNN-SVM algorithm has better performance than KSVM algorithm.

• 出版日期2008
• 单位南京林业大学