Enhancing Global Physical Activity Levels Through Personalized Sport Recommendations Using Machine Learning

EasyChair Preprint 15646

Abstract

Sedentary lifestyles are becoming the norm in the modern world. The development of obesity is increasing, which causes other health problems including cardiovascular diseases, musculoskeletal cases, and even diabetes mellitus. By providing tailored sports recommendations, people can remain active and minimize injury while reaching their goals. Relying on conventional approaches that do not take the interactions of other physical, medical, and psychological factors into consideration has resulted in bad sports decisions and associated problems.

To do so, we propose using a machine learning (ML) stacking ensemble model, which delivers personalized sports suggestions for the users depending on their individual traits. The model achieved 92% accuracy based on a custom dataset using demographic, physiological, and key activities/preferences data. Feature importance analysis identified key predictors, which included age, flexibility, endurance, and injury history. This study demonstrates ML's potential to overcome the limitations of traditional methods, contributing to safe and exciting participation in sports, as well as the disclosure of sports talents.

Keyphrases: Health and Fitness, Personalized Recommendations, Stacking Ensemble, machine learning

@booklet{EasyChair:15646,
  author    = {Sara Medetbekova and Zhenishbek Orozahunov and Amina Davidbek Kyzy},
  title     = {Enhancing Global Physical Activity Levels Through Personalized Sport Recommendations Using Machine Learning},
  howpublished = {EasyChair Preprint 15646},
  year      = {EasyChair, 2025}}