一种基于多模态模型的随机子空间分类集成算法

分类是当前机器学习的重要研究内容之一,已取得了一定的进展.现有的文本分类方法大多基于VSM模型,而VSM未能有效地利用隐含在文本中的结构信息.同时,VSM下的样本空间常常是高维的,单一的降维策略可能会丢失有用信息.为改进现有算法的不足,提出了一种基于多模态模型的随机子空间分类集成算法MMRFSEn,有效地利用文本中的结构信息（单词分布位置的均值和标准差）,且各基分类器是由随机选择的子空间构建而成.实验结果表明,该方法是有效可行的.

A Multi-modality-based Random Subspace Classifier Ensemble Algorithm

Text Classification is an important machine learning research, in which some progress has been made. Most of the existing classification methods are based on Vector Space Model（VSM） , but VSM does not effectively utilize the structure information hidden in the text samples. At the same time, VSM vectors are often high-dimensional, merelv using dimensionality reduction strategy may lead to the loss of the useful information. To uvercome the shortcomings of the existing algorithms, we propose an algorithm called Multi-modality-based Random Feature subspaee classifier Ensemble （MMRFSEn） . which can eft）etively use the structure information hidden in the text such as the words＇ s average localion and standard deviation, and meanwhile each single classifier is constructed by a randomly selected subspace. The experimental results show that the newly developed method is effective and feasible.