Abstract

The e-Health care system enables us to store patient’s personal health record online. Now a days, doctor’s decisions on health of patients is gaining importance in treating serious diseases. The overall health of human body can be subjected to many clinical parameters like random blood sugar level, white blood cell count etc. In addition to clinical parameters, the state of set of symptoms of all diseases contributes to overall well-being of a human being. Due to this the health of a human body can be decided by a set of parameters which include clinical parameters that decide the health of various organs in our body and symptoms associated with various diseases. Each of the clinical parameter can be associated with a reward based on its value being fallen in a particular bin. Also symptoms can be associated with a reward based on its intensity. The doctor will take many actions against a patient such as giving appropriate medication in course of tablets, operating surgeries, giving salination etc. So this system consists of set of clinical parameters and symptoms together as states in a model of machine learning. The set of actions taken by the doctor constitute actions of an agent where doctor is treated as an agent in this model. So a set of clinical parameters and symptoms were taken and a specified number of actions is taken to assess the performance of model in basic reinforcement learning learning and epsilon-greedy approach of machine learning. Results show that Q learning outperforms reinforcement learning and epsilon-greedy approach and these results enable the doctor for better decision making