多策略融合的SVM决策树构建

为了在提高文本分类效率和提升分类速度间进行平衡,综合考虑SVM决策树的深度、均衡度、构造方式、类内样本数、类间相似度等对分类结果的影响,提出针对海量文本多分类问题的SVM决策树构建算法.在大规模语料库上的文本分类实验表明,该算法可在一定程度上提升分类效果,同时可以大幅减少训练和测试时间,方法可行且适应性强.

Construction of SVM decision tree with fused multi-strategies

In order to make balance between the improvement of the text categorization efficiency and the promotion of the categorization speed,the influence of the depth balancing degree,construction mode inner sample number,and inter-category similarity of SVM decision tree on the categorization result was comprehensively considered and a construction algorithm of the SVM decision tree for massive text categorization was proposed.The experiment on text categorization of massive linguistic corpus demonstrated that this algorithm could improve the categorization efficiency to some extent and decrease greatly the training and testing time at the same time.This method was feasible with strong adaptability.