Hospital bed management support using regression data mining models

The lmitations found in hospital management are directly related to the lack of information and to an inadequate resource management. These aspects are crucial for the management of any organizational entity. This work proposes a Data Mining (DM) approach in order to identify relevant data about patients’ management to provide decision makers with important information to fundament their decisions. During this study it was developed 48 DM models. These models were able to make predictions in the hospital environment about beds tournover/patients discharges. The development of predictive models was conducted in a real environment with real data. In order to follow a guideline, the Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology was adopted. The techniques used were the Regression Tree (RT) and Support Vector Machine (SVM) in order to perform regression tasks. Regression models were able to predict patient’s discharges with Relative Absolute Error (RAE) lower than 100% ]38.26; 96.89[. Significant results were achieved when evaluated the Mean Absolute Error (MAE) ]0.619; 4.030[ and Mean Squared Error (MSE) ]0.989; 34.432[ .The use of these models can contribute to improve the hospital bed management because forecasting patient discharges makes possible to determine the number of beds available for the subsquent weeks.