Study On The Prediction Model Of Syndrome Elements For Diabetic Lower Extremity Arterial Disease Based On Decision Tree And Neural Network

Objective To construct the prediction model of syndrome elements for diabetic lower extremity arterial disease based on decision tree and neural network,and provide clinical assistance.Method In this study,we adopted a retrospective study method.600 inpatients in the Department of Endocrinology of the Affiliated Hospital of Shandong University of TCM from 2020 to 2022 meeting the research standards were collected,including300 patients with DLEAD and 300 patients without DLEAD.The information of gender,age,test results and four diagnoses of traditional Chinese medicine were collected,and the information was recorded into Excel to establish a database.SPSS21.0 was used for statistical analysis.The count data were described statistically by frequency and percentage,and analyzed statistically by Chi-square test.The measurement data were statistically described by mean ± standard deviation,and statistically analyzed by T-test.Syndroms with frequency ≥1% were selected to be included in the decision tree and neural network.CHAID,CRT and QUEST algorithms were used to build the decision tree model,and multi-layer perceptron and radial basis function algorithms were used to build the neural network model.Further,the prediction rules of each model were comprehensively analyzed to obtain the prediction rules of DLEAD that can guide clinical practice.Results1.There were no statistical differences in gender,age,course of disease and glycosylated hemoglobin between the two groups.2.In the group without DLEAD,the top 3 syndrome elements were qi deficiency(63.33%),yin deficiency(51.33%)and qi stagnation(26.67%).In the group with DLEAD,the top 3 syndrome elements were qi deficiency(71.67%),blood stasis(70.33%)and yin deficiency(57%).3.The decision tree model based on CHAID algorithm generates 7 recognition rules.Blood stasis,excess heat,water stagnation,blood deficiency and phlegm are selected into the model as recognition variables,and blood stasis is the best recognition variable.The prediction accuracy rate was 74.5%.4.The decision tree model based on CRT algorithm generates 10 recognition rules.Blood stasis,excess heat,blood deficiency,water stagnation and phlegm are selected into the model as recognition variables,and blood stasis is the best recognition variable.The prediction accuracy was 75.7%.5.The decision tree model based on QUEST algorithm generates four recognition rules.Blood stasis,water stagnation and excess heat are selected into the model as recognition variables,and blood stasis is the best recognition variable.The prediction accuracy rate was 72.5%.6.The top five standardization importance of the neural network model based on the multi-layer perceptron algorithm are blood stasis,water-stagnation,excess heat,phlegm and Yin,with a prediction accuracy of 77.8%.7.The top five standardization importance of neural network models based on radial basis function algorithm are blood stasis,phlegm,water stagnation,excess cold and excess heat,with the prediction accuracy of 75.6%.Conclusion1.The model constructed in this study has good predictive performance and good fitting effect,which can provide ideas for the differentiation and treatment of clinical DLEAD in Traditional Chinese Medicine,and has certain reference significance.The predictive performance of CRT algorithm decision tree is slightly better than that of CHAID and QUEST algorithm decision tree model,and the predictive performance of multi-layer perceptron algorithm neural network model is better than that of radial basis function algorithm neural network model.2.Blood stasis is the most valuable factor in predicting DLEAD.The optimal recognition variable of the decision tree models of the three algorithms constructed in this study is blood stasis,and the neural network models obtained by the two algorithms regard blood stasis as the one with the highest standardization importance.3.The risk of DLEAD is higher in patients with intermingled blood stasis and heat and patients with water stagnation.