This report presents a bed exit prediction model using a decision tree classifier, a machine learning method. This system is designed to alarm a caregiver in a nursing home to prevent bed­ falls among institutionalized elderly. For the input of this algorithm, features are derived from four force sensing resistor sensors and six piezoelectric sensors mounted on a device which is placed under the mattress at chest level. The classifier is based on data containing 98 bed exits of two different people on two different mattresses. The best results were obtained for the feature which uses the standard deviation of the force sensing resistor sensors of the last 10 seconds. The classifier was evaluated using a test set which contained a total of 645 time frames of 5 seconds of 23 cases where someone leaves the bed and 622 time frames where there is no bed exit. 10 windows with bed exits were classified correctly out of 23. Moreover, the system has a false alarm rate of 3.9%, meaning that in 24 time windows a bed exit is predicted while there is no actual bed exit. The accuracy with which the model classifies a bed exit or no bed exit correctly 5 seconds beforehand is 94.3%. Thereafter, it was researched whether adding an extra sensor plate would be helpful to improve the bed exit prediction. The extra plate should be placed on the side of the bed if the preferred side for getting up is known, or at the hip level if the side is unknown. This is of added value to define the bed status of the client as this plate, in contradiction to the sensor plate at chest level, still measures pressure when someone sits up straight.