The invention relates to a
patient evaluation classification method based on an improved LDA
topic model, and aims to efficiently and accurately classify patient written evaluations served by general practitioners. Firstly,
patient evaluation is processed through a series of preprocessing steps including word segmentation, abbreviation conversion, stop word removal, number and special character removal and low-frequency and high-frequency vocabulary removal. Thirdly, analyzing the data by applying various
unsupervised clustering algorithms to generate a plurality of clustering results, and calculating
ambiguity so as to construct a high-confidence
data set; and training an improved LDA
topic model by using the
data set, including calculating an inter-class
scatter matrix and an intra-class
scatter matrix, updating a projection matrix, re-clustering the
data set by using the projection matrix, and determining 60 class center topics. And finally, determining a text topic category for a given user evaluation text. The innovation of the invention lies in that multiple clustering algorithms and high-confidence
data selection are combined, the accuracy of
topic model training is improved, and through 60 finely divided topic categories, the
patient evaluation classification is more accurate, and the improvement of the medical
service quality is facilitated.