The social cost for the increasing older population is motivating new research, designed to support the ability and to favorite the autonomy of the disabled as well as the elderly. This paper describes an original proactive application of a Kohonen classification for home monitoring. Our approach attempts not to use any additional sensor and is only based on the observation of existing home automation events as non vectorial data. An analysis of user habits through the dissimilarity of services with a Kohonen algorithm is proposed for offering automatic scenarii in adequacy with the user capabilities. The simulation results, issued from real data from previous experiences, show the interest of our algorithm.