This study addresses the insufficient consideration of physical activity in managing Type 1 diabetes mellitus, using real patient data from the Ohio dataset to assess its impact. We utilize GRU models, known for capturing temporal dependencies, to analyze physical activity patterns and their influence on glycemic state. Additionally, we examine various parameters affecting GRU performance, such as input sequence length, hidden units, and learning rates, to understand how they impact accurate prediction of physical activity effects on glycemic state. By lever-aging the Ohio dataset and exploring different GRU parameters, we aim to improve personalized diabetes management strategies tailored to individual patient requirements.