Application of the Weka Machine Learning Library to Hospital Ward Occupancy Problems

Abstract

We explore the potential of applying machine learning techniques to the management of patient ow in hospitals. For this project, we have obtained the Weka machine learning library and three years of historical ward occupancy data from Rockyview Hospital. We use Weka's classifier algorithms and the Rockyview data to build a model of patient ow through each ward. Using Weka, we then attempt to predict ward occupancy problems on any given day using the model and the ward conditions from the previous day. This process is repeated for all eighteen wards. Finally, we obtain rules (sets of ward conditions that warn of an impending occupancy problem) for each ward and present the results.