We use decision tree induction algorithm to generate decision tree models for the ICDDR,B hospital surveillance data. Various preprocessing and formatting activities have been carried out on data at first to make data ready to build the models. Decision rules are generated from those models then. Finally the rules are used to classify patients into three classes: High, Mid, and Low, based on their critical condition so the hospital authority could take prudent actions on critical patients. We use different techniques to generate an optimal decision tree and compare the generated trees with different performance metrics, e.g., accuracy, precision, recall, etc.
展开▼
