Optimizing Calorie Prediction for Wearable Devices: A Machine Learning Approach with Simulated Fitness Data

This article attempts to explore the influencing factors and predictions of human calorie consumption through data visualization analysis and the use of machine learning methods, using a highly similar dataset collected by wearable devices for simulation. Firstly, based on the visualization tool Seaborn in Python, the relationship and distribution of various influencing factors are obtained through visual analysis; Then perform correlation tests on the dataset to identify the main factors affecting human calorie expenditure; Finally, the random forest model and XGBoost model were selected to predict calorie consumption, and the models were compared using three indicators: MeanAbsolute Error, MeanSquareError, and R-squared Score. The results showed that the optimized hyperparameter XGBoost (Tuned) model had better predictive performance for calorie consumption, with an MSE value of 1.461 and a MAE value of 0.882. This study can provide reference for personalized health management and the development of personalized exercise training programs.