Data Analytics Based On Propensity Score Methods And Its Application In Health Care

With the development of information technology, large volumes of data are generating every day in different fields. How one can discover knowledge from data has become research focus. One of those important research problems in data-driven study is causal inference between intervention and result from observational study. Observational data from observational study are influenced by selective bias, and research in data analytics should focus on keeping the data balance and removing the selective bias. Propensity score method is a data analytics method, and it can be used to remove the influence from selective bias to casual inference.Based on the demand in the field of health care, such as treatment plan and healthcare policy, we started the study on propensity-score-based data analytics methods. The algorithm based on propensity score methods was provided. Monte Carlo simulation was used to evaluate the performance of the algorithm. The result of simulation suggests that the algorithm can remove selective bias. PSB data analytics platform was designed and developed to support observational study and propensity score method analysis. It could act as a tool for data analysis. At last, based on a case study in health care, PSB data analytics platform is displayed.This thesis firstly illustrates the difficulty, practical demand and current state of research in causal inference from observational study and data analytics. The second chapter introduces the basic concepts and methods of propensity score methods. The algorithm of acquiring and matching propensity score is given. Monte Carlo simulation was used to evaluate and compare the performance of the algorithm. The third chapter gives the design plan of PSB data analytics platform. The main focus is on the application-oriented platform configuration, platform framework and some key functions. In the fourth chapter, we demonstrate how to use Java and R language to develop the system. The main focus is on data management, data mapping and realization of algorithm calculation. The fifth chapter shows the case study through the PSB data analytics platform.This thesis’s research focuses on the application of propensity score methods in observational study, and the development of PSB data analytics platform to support observational study and data analytics. Further research could explore more robust properties of propensity score methods and the improvement of data analytics platform.