Utilizing econometric/A.I. model to analyze the change in individuals’ behaviors in medical and advertising data

Abstract

This thesis analyzes individuals’ behavioral changes using econometric and artificial intelligence models for medical and advertising data. These researches provide econometric implications and present directions for ways to utilize medical and advertising data. The thesis consists of three papers. The first paper analyzed how patients' hospital visits change with the spread and resolution of international infectious diseases through medical data. Through difference-in-difference analysis, it was confirmed that in the case of patients suffering from mild diseases, visits decreased during the diffusion period and increased after the resolution. In addition, it was found that the increase or decrease of visits differed according to the type of hospital and individual characteristics. The second paper verified the efficiency of advertising repetition strategies and variation strategies with a survival model by utilizing the pre-roll video advertisement data. Groups were divided according to the length users watched the pre-roll ad when first exposed to the specific advertisement brand. Then the study analyzed how each group responded differently to repetition strategies and variation strategies. Accordingly, it was confirmed that the variation strategy was more efficient in the group hostile to the advertisement than the group favorable to the advertisement and that they also responded differently to the repetition strategy. The last paper uses two-part and deep learning models to analyze and predict a radical increase in end-of-life healthcare expenditure. The empirical analysis shows that past healthcare expenditures as an explanatory variable could successfully replace the proximity to death, which causes a problem when forecasting since proximity to death is not observable.