Construction of a predictive model for fear of fall in rehabilitating elderly stroke patients using a multi- layer perceptron neural network

Abstract Objective To establish a predictive model using multi-layer perceptron (MLP) for fear of fall in elderly stroke patients during the rehabilitation period. Methods From June 2022 to February 2023,368 elderly patients with rehabilitation stroke were investigated by scales.Conduct univariate and multivariate analysis of the influencing factors for fear of fall, using multivariate Logistic regression and MLP to establish the prediction model and calculate the prediction accuracy of the two models.Predictive efficacy was assessed using the receiver operating characteristic (ROC) curve. Results The prediction accuracy of the multivariate Logistic regression model was 78.00% and the area under the ROC curve was 0.848; the prediction accuracy of the MLP model was 84.90% and the area under the ROC curve was 0.890. Conclusion The prediction of fear of fall in elderly stroke patients during the rehabilitation period can be done with MLP model.