基于优化密度的耦合空间LDA文本聚类算法研究

Coupling space LDA text clustering algorithm research based on optimizing density

针对传统的空间向量模型在进行文本表示时计算相似度仅采用词频统计来表示文本以及对高维文本数据聚类效果有所下降等问题,提出一种基于优化密度的耦合空间LDA文本聚类算法。该算法利用提出的耦合空间模型和LDA主题模型线性融合计算文本相似度,并对阈值敏感问题进行优化,确定不同密度区域对应的阈值半径。实验结果表明,与改进的DBSCAN文本聚类算法和R-DBSCAN文本聚类算法相比,该算法的文本聚类精度更高、聚类效果更优。

Aiming at the problem that traditional vector space model to calculate the similarity in text representation only use statistic the frequency of the word to represent text and to the high-dimensional effect decreased of text data clustering, the paper proposed a coupling space LDA text clustering algorithm based on optimizing density. Linear fusion coupling space model and LDA theme model for computing text similarity, and optimized the issue of threshold of sensitive, the radius of threshold corresponding to the different density area. Experimental results show that, comparing with the improved DBSCAN text clustering algorithms and R-DBSCAN text clustering algorithm, the proposed algorithm performs higher accuracy and better clustering effect in text clustering.