Deep learning: How computer scientists are using fitness tracking data for workout recommendations

An image to demonstrate fitness tracking, which deep learning tools could assist with by giving workout recommendations.
© iStock/Todor Tsvetkov

A new deep learning which uses fitness tracking data for workout recommendations and predictions has been developed by computer scientists at the University of California San Diego.

The tool developed for workout recommendations, FitRec, has been trained on a data set of over 250,000 fitness tracking records for over a thousand runners. FitRec analyses past performance to predict speed and heart rate given specific future workout times and routes.

FitRec can also:

  • Identify important features that may affect performance, for example, a hilly route or the user’s level of fitness;
  •  Recommend alternate routes for runners who want to achieve a specific target heart rate; and
  • Make short term predictions, for example, advising runners when to slow down to avoid exceeding their desired maximum heart rate.

Deep learning from fitness tracking data

Deep learning is a type of artificial intelligence, and a method of machine learning which utilises artificial neural networks.

For the purpose of personalising workout recommendations, a tool which uses all of the fitness tracking data to learn, but also can learn personalized dynamics from a small number of data points per user, was required.

The researchers used a type of deep learning architecture called long short-term memory networks (LSTM), which they adapted.

Personalised workout recommendations

Julian McAuley, a professor in the Department of Computer Science and Engineering at UC San Diego, commented: “Personalisation is crucial in models of fitness data because individuals vary widely in many areas, including heart rate and ability to adapt to different exercises.”

The researchers explained: “The main challenge in building this type of model is that the dynamics of heart rates as people exercise are incredibly complex, requiring sophisticated techniques to model.”

According to the University of San Diego California, FitRec could be trained to use other data in the future. For example, the way a user’s fitness evolves over time, or more complex recommendation routes such as safety-aware routes. However, before the tool could be used in commercial fitness apps, researchers would require more detailed fitness track data and deal with data quality issues.

Laboratory Supplies Directory - Now Live

LEAVE A REPLY

Please enter your comment!
Please enter your name here