Early Warning and Screening of Elderly Cognitive Impairment Based on Machine Learning Algorithm

Abstract

In order to improve the classification accuracy in the early warning and screening process of elderly cognitive impairment, this paper constructs a screening system for elderly cognitive impairment based on the survey data of community elderly residents in Changsha. The cognitive level of all samples was divided into normal, mild cognitive impairment (MCI) and cognitive disorder. Firstly, the correlation between all features and sample categories is described by mutual information, and the features that have no significant impact on sample classification are eliminated. Secondly, support vector machine (SVM) and random forest were used for sample classification. When determining the hyperparameters of the model, the learning curve based on generalization error is used for parameter combination optimization, and a variety of evaluation indexes are used to evaluate the performance of the model. Experimental results show that SVM has more accurate classification ability than random forest, while random forest is more “conservative” and tends to identify normal samples as abnormal ones, which can reduce the risk of loss of potential patients and is more suitable for the situation where screening work needs to find potential patients as much as possible.