Diseases Prediction from Officially Anonymized Medical and Healthcare Big Data

With the advanced technologies for ubiquitous computing, the large sensitive big data is valuable to explore the associations of data for epidemiology. New data privacy regulations allow sharing anonymization of data for relatively low cost than that of the conventional cohort studies. In this paper, We analyze the disease risk using Anonymously Processed Information to predict the diseases risk using the anonymously processed information of medical examination and health insurance claim data consisting of 200,000 and 290,000 individuals. We aim to demonstrate that using Anonymously Processed Information should be an additional methodology to accompany conventional cohort data analysis. We analyze a diseases risk by performing a logistic regression and calculating the relative risk, and develop some machine-learning models that predict the likelihood of a disease in three years, given an individual medical examination. Finally, we reveal the reduction in accuracy of the analysis as a result of
$k$
-anonymizing data.