论文部分内容阅读
针对具有多观测样本的相似不完整数据分类问题,提出基于SVM和多观测样本的相似数据分类算法.每类数据的多观测样本集由属于同一模式的单观测样本组成,每次分类时,对两个多观测样本集的标签做两次假设,通过比较不同标签假设下的分类误差确定多观测样本集的标签.该方法同时充分利用了样本类内的相关性和类间的差异性,实现了相似不完整数据的分类.实验结果验证了所提出方法的有效性.
Aiming at the problem of similar incomplete data classification with multi-observation samples, a similar data classification algorithm based on SVM and multi-observation samples is proposed. The multi-observation sample set of each type of data consists of single observation samples belonging to the same pattern. For each classification, Two hypotheses were made on the labels of two multi-observation sample sets, and the labels of multiple observation sample sets were determined by comparing the classification errors under different label assumptions. This method also took full advantage of the correlations and the differences among classes in the sample classes The classification of similar incomplete data is given.The experimental results verify the effectiveness of the proposed method.