Comparison Of Logistic Regression Model,Neural Network Model And Decision Tree Model In Prediction Of Mild Cognitive Impairment To Alzheimer's Disease

Objective To understand the outcomes of mild cognitive impairment in Alzheimer's disease and to explore the main risk factors for the decline of mild cognitive impairment to Alzheimer's disease;to establish mild cognitive impairment to Alzheimer's disease The prediction of the prediction of the prognosis of Alzheimer's disease is based on the prediction of the prediction of the mathematical model of the syndrome and the accuracy of different models.Methods A total of 428 MCI patients were collected in 5 communities in Nanchang City.A total of 17 variables including demographic characteristics,lifestyle factors,past medical history,ADL scale,MoCA scale,and urinary AD7c-NTP protein were collected.The collected case data was used to establish the database and data entry using EPidata3.1 software;the SPSS software was used to establish the prediction model using Logistic regression,BP neural network model and decision tree model,and the area under the ROC curve was calculated using Medcalc software.Compare the prediction accuracy of the three models.Results Of the 428 MCI patients,121 were found to be AD and the outcome rate was 28.3%.Multivariate logistic regression analysis showed that of the 17 variables studied,age(2)(3)(OR=2.326,95%CI=1.699,3.014;OR=3.651,95%CI=2.241,4.537),gender(OR=2.603,95%CI=1.626-4.810),smoking(OR=1.157,95%CI=1.073-1.341),heavy drinking(OR=1.157,95%CI=1.026-3.587),MocA(OR=0.766,95%CI= 0.681-0.861),ADL score(OR=1.790,95%CI=1.637-1.979)with MCI outcomes to AD was statistically significant.Logistic predictive model was P=1/(1+e(1.885+1.537×age(2)1.825×age(3)+1.687×gender+1.249smoking+1.374×heavy drinking-0.267× MocA score +1.318×ADL score)).The area under the ROC curve is 0.827(95% CI: 0.789-0.855)under the ROC curve;the importance of the predictors of the BP neural network model is ADL score,MoCA score,age,urine in descending order of importance AD7c-NTP,gender,drinking,smoking,and whether to participate in physical exercise or sports labor,AD-family history.The area under the ROC curve for the AUC was 0.819(95% CI 0.779 to 0.848,p<0.0001).Decision tree model predictor variables include age,ADL score,AD-family history,AD7C-NTP,drink,MoCA score,gender,whether to participate in physical exercise or sports labor,the area under the ROC curve of the predictive model is 0.861(95% CI: 0.832-0.881);the ROC curves of the three models The comparison of the area under BP neural network(P=0.002),logistic regression model(P=0.003),and ROC area value of the decision tree model showed statistically significant differences.However,there was no statistical difference in the area under the ROC curve between BP neural network and Logistic regression(P=0.462).Conclusion The rate of conversion of MCI patients to AD was higher.The risk factors for outcomes were older,women,smoking,heavy drinking,and ADL scores;protective factors were high MoCA scores.The decision tree model is superior to the BP neural network and Logistic regression model in predicting MCI outcomes for AD performance.