Technological enhancements have advanced research in smart homes and intelligent environments. Because this research area is quickly maturing, potential users are investigating this technology as a means of supporting the independent life of older people and people with disabilities. Numerous intelligent devices are now available, which provide the resident with 24-hour health monitoring. In this paper we go one step further with the idea of pairing learning and prediction to such devices and making them adaptable to residents. We compare the performance of alternative learning algorithms for predicting health data collected from these intelligent devices in a smart home. Finally, we evaluate the learning algorithm performance in predicting abnormality using the collected data and everyday activities in a smart home. If successful, this capability will give these intelligent systems an additional advantage of prediction which would help the residents to take relevant precautions.