Leisure-time physical activity (LTPA) has been shown to be an effective way of preventing diseases. However 50% of the people who start any type of sports or LTPA drop out of the program within 6 months. Even though a lot of research on exercise adherence has been performed, no common consensus on all the leading factors to adherence exists. Cultural differences make studies in other countries not suitable for Japan. In this paper we present designs of recommender systems to recommend LTPA to adults. Due to complexities of problem domain and vast number of possible LTPA, first we assume people with similar lifestyle may follow each other better than others as far as exercise is concerned. Based on lifestyle data in complete health check-ups (Ningendoku) that is performed in Japan, we find similar people. In addition, based on expertise and detailed information of different sports and activities we build ontology trees and tables for different attributes of LTPA, these ontologies are used to calculate distances of different LTPA. Health-care professionals in preliminary discussion showed a lot of interest on our system.