Effect of Physical Fitness on Body Mass Index through Artificial Intelligence in South Chinese College Students from 2013 to 2025
This study analyzes data from over 83,000 Chinese college students from 2013 to 2025 to demonstrate that physical fitness indicators are strongly associated with Body Mass Index and that an XGBoost machine learning model can effectively predict BMI, with pull-ups, vital capacity, and long-distance running identified as key predictive features.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
Imagine your body as a high-performance car. For years, mechanics have known that if the engine sputters (poor fitness) or the car is weighed down by too much cargo (excess body fat), the vehicle won't run smoothly. But what if you could predict exactly how heavy that cargo is just by listening to the engine, checking the tires, and seeing how fast the car accelerates? This is the heart of a new study that sits at the intersection of physical education and computer science. It asks a simple question: Can we use a student's performance on standard gym tests to guess their Body Mass Index (BMI)? BMI is a common way to measure if a person is underweight, healthy, overweight, or obese, calculated by comparing their weight to their height. While we know being heavy makes it harder to run or jump, the exact recipe of how different fitness skills mix together to reveal a person's body composition has been a bit of a mystery. Now, scientists are using "Artificial Intelligence" (AI)—think of it as a super-smart digital detective that can spot hidden patterns in massive piles of data—to solve this puzzle.
This study, conducted by researchers from Shenzhen University and Louisiana State University, decided to let the computer detective do the heavy lifting. They gathered data from a massive group of 83,664 first-year college students in Southern China, spanning from 2013 to 2025. That's like filling a giant stadium with students and asking them to take the same six physical tests: blowing into a tube to measure lung power, jumping forward from a standstill, stretching to touch their toes, sprinting 50 meters, running a long distance (800 meters for girls, 1,000 for boys), and doing either pull-ups (for boys) or sit-ups (for girls). The researchers then fed these results into four different AI models to see which one could best guess the students' BMI categories.
The results were like finding a golden key. The AI didn't just guess; it learned. The best model, a sophisticated algorithm called XGBoost, managed to predict a student's BMI category with about 75% accuracy for boys and 83% accuracy for girls. This suggests that physical fitness and body weight are tightly linked, like two gears turning together. But the real magic was seeing which tests mattered most. For the male students, the pull-up test was the superstar, acting as the single most important clue. It accounted for nearly half (45.27%) of the AI's ability to make a correct guess. This makes sense because a pull-up is a "weight-bearing" exercise; you have to lift your entire body against gravity. If you are carrying extra weight, it becomes significantly harder to do them, so the number of pull-ups a boy can do is a huge hint about his body mass.
For the female students, the story was a bit more like a team effort where no single player dominated. Instead of one test taking the lead, the AI found that lung capacity (vital capacity) and the long-distance run were the top detectives, working together with the other tests to figure out the BMI. Interestingly, the sit-up test, which measures core strength, wasn't as big of a clue for girls as the pull-up was for boys. The researchers suggest this might be because sit-ups focus on a specific muscle group rather than moving the whole body's weight, and most of the girls in the study were already in a healthy weight range, making it harder for the AI to spot differences.
The study also looked at how different types of running mattered. The long-distance run was a much stronger clue for BMI than the short 50-meter sprint. Think of it this way: sprinting is like a quick burst of explosive energy, which doesn't care as much about how heavy you are. But running for a long time is like carrying a heavy backpack for miles; the heavier you are, the more it slows you down. The AI picked up on this, showing that endurance is a better indicator of body composition than pure speed.
While the AI was very good at guessing, it wasn't perfect. It was great at identifying students with a "normal" weight, but it sometimes struggled to spot the very few students who were underweight or very obese. This is likely because there were simply far more healthy-weight students in the data, making it harder for the computer to learn the patterns for the smaller groups. The researchers are careful to note that this study shows a strong connection, or association, between fitness and weight, but it doesn't prove that one causes the other. It's like seeing that clouds and rain often happen together; the AI sees the pattern, but it doesn't tell us which one started the storm.
Ultimately, this research suggests that the routine physical tests students already take in school are more than just grades on a report card. They are powerful, practical tools. If a student struggles with pull-ups or long-distance running, it might be a friendly, early warning sign that their body composition needs attention. By using these simple tests and letting AI help interpret the results, schools could potentially screen for weight issues early and encourage students to build strength and endurance, keeping their "cars" running smoothly for years to come.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.