A Martial Arts Nutrient Recommendation System by Multi Regressor Model

Abstract

The given thesis offers an AI-based martial arts nutrition prescription system as the solution to the lack of accessibility to personalized dietary advice. Conventional nutrition education is based on direct training forming physical and resources boundaries. In this study, the project will create the prototype of the first mobile nutrition system in martial arts practice, making professional nutrition counselling much more accessible to people anywhere. The study uses the multi-output Random Forest regression model to foresee the individual macronutrient requirements, basing them on demographic factors about the user, training objectives, and activity levels. The system consists of Flutter mobile app, Firebase backends, and Python machine learning pipelines. Data augmentation and metabolic calculations through a 35-feature engineering pipeline expanded the dataset (231) to more than 1000 samples. The model produced outstanding results, characterized by R 2 = 0.961, MAE = 3.72g and MAPE 2.6%. The productions system covers input validation, intervals of confidence, meal planning, and response times less than seconds. Scientific accuracy and practical applicability were subjected to the approval of experts. Further development directions are based on exercise suggestions, smart device connection, culturally ruled preferences on food, performance in-depth metrics, and clinical system integration, making an all-in-one AI-enabled sports advice framework.