GenAI Enabled Workout Recommendation Using Temporal Learning Approach for Personalized Mobile Fitness Applications

Mobile fitness apps, popularized in the pandemic by rising health concerns, ease of use and flexibility, use wearable sensors to visualize user engagement for physical health, yet focus on generic rather than personalized health recommendations. To overcome this challenge, this paper proposes recommendation system centred around personalized physical activity suggestions, aims to enhance user motivation and encourage adherence to workout routines utilizing Generative Artificial Intelligence (GenAI) deep collaborative filtering. The solution offered here treats the recommendation problem as a sparse value prediction problem where user’s missing future workout duration is predicted using GenAI models that are able to create multiple workout combinations based on the user profile. In order to improve the accuracy of the prediction, latent representations created from explicit user data with domain-specific side features, are processed using generative transformers to produce highly tailored workout plans. An ablation study is conducted to measure how these components aid in personalisation, particularly novel workout combinations not included in the training data at all are, which the system produces, and are claimed in the training data are provided. This research aims to create more intelligent fitness recommendation systems using GenAI and accurately recommend activities to sustain physical activity while improving health issues, engaging users.