In Work (Reading, Mass.)
BACKGROUND : The health risk assessment aims to describe and evaluate the possibility of a certain disease, hospitalization, or death. With the in-depth research of big data and machine learning technology, the health risk of individuals can be assessed by using the technology, and intervention measures can be taken in advance to reduce the risk.
OBJECTIVE : This study aims to accurately predict and evaluate the possible risks of the population and individuals caused by environmental factors, and constantly improve the medical implementation process.
METHODS : The relationship between air pollutants and health risk is analyzed from three dimensions of the respiratory system, circulatory system, and digestive system, the prediction method of health quantity related to environmental factors is explored, and a hybrid time series model HTSM (Heuristic Test Strategy Model) based on nonparametric regression and residual fitting is proposed.
RESULTS : Respiratory and circulatory diseases are pollutant-sensitive diseases, while the elderly (> 65 years old) are the high-risk population. The improved model can effectively predict the unplanned readmission data in the actual medical scene, and the accuracy of the improved model is 11.11%higher than that of the traditional prediction model. In contrast to the single prediction model, HTSM's error index for different systems is much lower. The mixed model HTSM is better than the single model in fitting the original data.
CONCLUSION : HTSM model based on time series can effectively predict pollutant-sensitive diseases, which can provide an effective theoretical basis for assessing and predicting the population and individual health risks.
Ma Wangrong, Jin Maozhu, Zhen Weili
Machine learning, cost sensitive learning, environmental factors, health risk assessment, time series analysis and prediction