Smart Home resident may be an Alzheimer patient needing continuous assistance and care giving. Because of forgetfulness, this person may realize activities of daily living erroneously. In order to assist this person automatically in Smart Home, all his performed actions and activities are observed through the embedded sensors of Smart Home, and applying the data mining techniques his activities are analyzed. Then information about his activities is provided and in the consequence, comparing learned correct patterns and current observations the Smart Home may infer provision of assistance to this person at the appropriate moment. In this paper we propose a data-driven activity modeling approach, which supports reasoning in correct realization of the activities. Activities are presumed as the series of fuzzy events that occur shortly one after another. Per each activity, we calculate a fuzzy conceptual structure, and the model of activity is represented in form of a multivariable problem.