基于粒计算的哈夫曼树SVM多分类模型研究

针对多分类问题,将粒计算与最优二叉树相结合来构建SVM多分类模型。应用粒计算思想粒化多分类问题,计算出每个类别的粒度；以粒度为权值集合,构建哈夫曼树,以解决类内样本分布不均和分类效率低下的问题；对粗粒结点分别设计多个SVM分类器；最后,以低温存储罐材料多分类问题为研究背景,对模型进行了仿真验证。与其他方法的对比分析表明,该模型提高了分类效率,为多分类问题的处理提供了一个新的研究思路。

Research of SVM Multiclass Model Based on Granular Computing & Huffman Tree

In view of the multi-classification problems,we built the SVM multiclass model based on granular computing and Huffman-tree.After applying granular computing to grain classification problem,we could calculate the granularity and build the Huffman tree based on granularity weight set,which solves the uneven distribution of samples in the class and lows classification efficiency.We also designed SVM classifier for coarse grain nodes,and selected the low temperature storage tanks material multi-classification problem as the research background to simulate our model.Meanwhile,we compared our model with other methods.The result shows that the new model improves the efficiency of classification.It provides a new idea and a perfect method for multi-classification problem.