RecFit: a context-aware system for recommending physical activities

Abstract

Many people are bored with their current physical activities and would like individualized recommendations of alternatives. Even users who have favorite exercises may seek recommendations if their context (e.g., bad weather, location) changes. Prior work has focused on tracking user activities and goal-setting, but not on recommendations. In this paper, we describe RecFit, which systematically suggests physical activities based on the user's context (e.g. risk tolerance, budget, location, weather). RecFit works from 137 activities selected from the 2011 compendium of physical activities in order to recommend the 5 most suitable recommendations for each user. We describe our filtering criteria, algorithms, prototype and RecFit's activity database, which augments activities with metadata of ideal performance context (popularity, sociability, risk, location, expense, time, and weather).