Research on Cluster Analysis of Tibetan Medicine Syndromes type of the plateau common disease

The plateau common disease (chronic atrophic gastritis) is a digestive tract disease with typical Tibetan characteristics. At present, the unclear type of the syndromes hinder the further study of the plateau common disease (chronic atrophic gastritis). In order to weaken the previous subjective experience, from the point of view of machine learning, this paper uses clustering algorithms in data mining to classify them objectively, and combines clinical diagnosis and treatment data to put forward the research ideas of Tibetan syndromes type classification. From the point of machine learning, cluster algorithms in data mining were used in this paper to divide the syndromes type. The train of thought of Tibetan medicine syndromes type classification research was proposed, which combined with clinical diagnosis and treatment data. Firstly, the two step clustering and K-means algorithm were used to carry out as a preliminary result for the syndromes type of the plateau common disease (chronic atrophic gastritis). According to the different clustering results, the accuracy of the four classical classification algorithms is compared, and the optimal number of clusters is initially determined. Then, aiming at the characteristics of the data set and the performance mechanism of the above algorithm, the Gower's metric + improved K-Modes cluster method was proposed, and the plateau common disease (chronic atrophic gastritis) was divided into four syndrome types by the R language implementation. It can not only classify the plateau common disease (chronic atrophic gastritis) from a scientific point of view, but also can greatly improve the objectivity, standardization and accuracy of the syndromes type of Tibetan Medicine. Finally, based on the results of Gower's metric+ improved K-Modes cluster analysis, the symptom characteristics of each type of syndrome were summed up through the analysis of symptom frequency. The accuracy of prediction was 79%, which compared with the expert's experience.