【摘要】 本文研究了PACS中诊断文本的聚类问题。对诊断文本通过词频矩阵建立向量空间模型,在保持文本向量余弦相似度不变的前提下,采用奇异值分解对空间降维,以减少空间的占用和计算时间,最后利用层次聚类算法得到各种常见病变的诊断文本聚类结果。最后从实践角度讨论诊断文本聚类需要考虑的因素。更多还原

【Abstract】 In this paper, we use text clustering techniques to solve the diagnosis text cluster. After constructing the term-document matrix, each diagnosis text is represented as a dot in vector space. As there are many texts, singular value decomposition is adopted to reduce the high dimensional space. Hierarchical clustering algorithm, a simple algorithm is used to cluster the diagnostic text. Implementation considerations are included in the end.更多还原