Desenvolvimento e avaliação de um sistema web para geração de planos de treino com a API do ChatGPT

Abstract

With the increasing number of students in gyms, the demand for Personal Trainers to develop personalized workout plans has grown significantly. However, manually creating these plans requires considerable time, making it difficult to personalize and continuously monitor each student. This study proposes the development of a system based on large-scale language models for the automated generation of workout plans using the ChatGPT API. Different prompt engineering approaches are analyzed to evaluate combinations of techniques that generate more precise and tailored recommendations according to the proposed evaluation criteria. Experiments conducted on a diverse dataset of students demonstrate that combining the Persona and Few-Shot techniques results in structured workout plans that are better aligned with training objectives.