Advanced Utilization of Medical Record Data with Relational Probabilistic Models

在这个研究项目中，我们试图为日常医疗过程中产生的病历数据开发一个新的统计分析框架。我们提出使用马尔可夫决策过程对病历数据建模，并使用强化学习和反强化学习方法评估患者状态和医生行为的价值。我们还开发了一个工具链，可以将病历数据可视化，提取数据的有趣部分，并使用提出的方法分析提取的数据。

In this research project, we tried to develop a novel statistical analysis framework for medical record data which are generated in daily medical care processes. We proposed to use Markov decision process to model the medical record data and evaluated the value of patient states and doctors' actions with reinforcement learning and inverse-reinforcement learning methods. We also developed a tool chains which enable to visualize medical record data, extract interesting parts of the data, and analyze the extracted data with the proposed methods.