Prediction of Risk Factors for Fall using Bayesian Networks with Partial Health Information

Abstract

Falls are frequent in the elderly and it is the number one cause of traumatic death in this population. Fall prevention requires to evaluate which risk factors for fall are present for a person on the basis of incomplete health data. In this context, the main objective of this paper is to identify the risk factors for falls in elderly people and to evaluate them based on partial observations. Health data for this study was provided by the hospital of Lille, France. Firstly, the risk factors are identified from the data with the help of an ontology of the risk factors for fall. Furthermore, the steps of data pre-processing, missing value imputation and variable selection are described. Lastly, risk factors are predicted using a Bayesian Network (BN). To evaluate the quality of the BN, we compared it with Logistic Regression, Decision Tree, Random Forest and Support Vector Machine. The result shows that no classifier is clearly better than the others and DT is worse than the baseline classifier. We have explained the interest of using BN to predict target variables on the basis of partial observations from a random set of variables.