基于加权模糊隶属度的二叉树多分类算法

针对二叉树支持向量机多分类算法准确率与分类效率较低的问题，提出了一种基于加权模糊隶属度的二叉树支持向量机多分类算法（binary tree support vector machines multi-classification algorithm based on weighted fuzzy membership，PF-BTSVM）。该算法依据最大最小样本距离与质心距离构造出一个近似完全二叉树，提高了整体结构的分类效率；利用模糊隶属度函数以及正负辅助惩罚因子对训练集进行筛选，剔除掉对分类无用的样本与噪声值，实现了训练集的提纯并且削弱了不平衡分类时超平面的偏移。在数据集上的实验结果表明，与其他二叉树多分类算法相比，该算法在提高了分类准确率以及稳定性的的同时还加快了训练与分类的速度，而且这种优势当分类的不平衡度越大时越明显。

Binary tree multi-classification algorithm based on weighted fuzzy membership

In view of the low accuracy and classification efficiency of binary tree support vector machine multi-classification algorithm, this paper proposed a binary tree fuzzy multi-classification algorithm based on auxiliary punishment factor(PF-BTSVM). The algorithm constructed an approximate complete binary tree based on the maximum and minimum sample distance and centroid distance, which improved the classification efficiency of the whole structure. It used the fuzzy membership function and the positive and negative auxiliary penalty factors to screen the training set, eliminated the useless and noise samples for the classification, which improved the training set and weakened the hyperplane offset in the unbalanced classification. Experimental results on data sets show that compare with other binary tree multi-classification algorithms, the proposed algorithm not only improves the classification accuracy and stability, but also speeds up the training and classification, and this advantage is more obvious when the degree of imbalance in classification is larger.