Digital Debunking: Can Machine Learning Help Design Better Workout Programs?
Have you ever walked into a local gym, eager to start your workout routine only to realize you’re overwhelmed and unsure where to begin? This can be uncomfortable and embarrassing, but it’s completely natural. Making the right choices for your body can be complex, especially if you’re unsure where to start. And a crowded gym where all the treadmills and machines are full only complicates things further. But even working out at home can be challenging; even the latest, trendiest fitness machine can go to waste if you’re not using it properly or consistently. The bottom line for most of us is this: without a solid plan in place, you won’t achieve the results you’re so eager to see.
In the absence of a personal trainer, look to technology as an alternative. Advanced machine learning and artificial intelligence (AI) can help you intelligently guide your exercise decision-making. Machine learning and AI have become integral parts of our daily lives, shaping how we interact with technology, make decisions, and access information. This got our team thinking: can machine learning design a workout program that meets the demands of average individuals starting their fitness journey? To find out, we turned to the solutions within the Altair® RapidMiner® data analytics and AI platform.
Designing a Reliable Exercise Program with Machine Learning
To answer the question of whether machine learning, specifically decision trees, could optimize the design of training programs for non-professional gym users, our team examined a generated dataset of 3,350 gym members . This dataset was the foundation for our decision tree model. A decision tree is a simple tool that clarifies decision-making criteria, applying consistent logic to reach desired outcomes. It ensures a more systematic approach to sport science.
We used Monte Carlo simulation – a type of simulation that uses randomness to solve deterministic problems – to generate the dataset, which contained two key variables:
- The average duration (measured in minutes) gym members spent on different exercise machines.
- The number of sets and the weight/resistance used for each machine.
This process allowed us to estimate the average outcome by running several simulations and averaging their results. While everybody’s goals differ, generally, gym members aim to achieve one or more of the following objectives: weight loss, improved endurance, improved strength, and overall health improvement. The dataset included variables representing the results of these four objectives after a six-month period (Figure 1).
Figure 1: Results of achieving set targets (over six months).
The workflow of using machine learning to optimize workout routines is as follows:
- Machine learning learns important exercises that lead to achieving objectives.
- Decision trees explain the nature of these exercises and guide individuals through an optimized exercise program (Figure 2).
By analyzing this data, you can identify the exercises needed to build an optimal exercise program.
Figure 2: Depiction of workflow process using machine learning and decision trees.
The process starts by joining the raw data of the average gym session duration and number of sets (Figure 3) with members' results after six months (Figure 4). This data is then used to develop four decision tree models to predict and explain the successful groups for each training objective. The decision tree models generate two sets of results identifying the groups with a high success rate and recognizing the critical variables that helped them achieve their stated objective.
Figure 3: Raw data of average weekly gym duration and sets for each exercise/machine.
Figure 4: Augmented data after six months combining exercise information with member results.
For instance, if a user was interested in increasing their endurance to improve marathon results, or enhancing strength to reach various weightlifting goals, decision trees can provide a roadmap to help them reach these goals. The following decision trees show the exercises needed to build an exercise program targeting the four original objectives: weight loss, improved endurance, improved strength, and overall health improvement. (Figures 5-8).
Figure 5: Decision tree model for weight loss.
Figure 6: Decision tree model for improved endurance.
Figure 7: Decision tree model for building strength.
Figure 8: Decision tree model for overall better health.
Practical Use of Results
By using the results from decision trees, individuals can create a training program design matrix: a tool that can help users, personal trainers, and coaches develop exercise programs. Decision trees change the way we look at fitness, providing a more methodical approach. The models' results provide the critical exercises and components needed to achieve one’s goals. The matrix below (Figure 9) is the result of the gym member data. For example, if you’re trying to lose weight, the model recommends a combination of cardio and weights. Incorporating running on the treadmill and swimming with squats and dumbbells is the most effective combination that can help you do this. Alternatively, if you’re trying to build strength, the decision tree model recommends moving towards free weights and the cable machine.
Figure 9: Training program design matrix results.
The greatest advantage of using these models is that they remove the guesswork from creating workout routines. By relying on data-driven insights, you can focus on becoming fitter and more active without feeling intimidated or overwhelmed. It can be a struggle when making fitness-related decisions. Human emotions, biases, and incomplete information – or too much contradictory information – complicate the decision-making process.
Thankfully, we now have a powerful tool that can help us better reach our fitness goals – no matter what those goals are. So, the answer to our question is confirmed – yes, machine learning and decision trees can build reliable roadmaps that can help you build informed, data-driven exercise routines.
Visit https://altair.com/altair-rapidminer to learn more about the Altair RapidMiner platform and its decision tree capabilities.
