The modeling of fuel flow in aircraft is a critical element in evaluating their performance. However, many of the existing approaches rely on highly aggregated data, which may not provide the necessary accuracy required by the aviation industry. To address this challenge, a new data-driven approach is proposed in this study. This method leverages machine-learning techniques and uses full-flight data from aircraft sensors to develop fuel flow models. The approach focuses on identifying the features that impact fuel flow rates during different flight phases, using a unique deep-learning approach for modeling. The results demonstrate a significant improvement over traditional practices, providing a more comprehensive and accurate alternative for performance evaluations. As a result, airlines and maintenance providers can optimize their operations and reduce fuel costs while meeting environmental targets. The proposed method achieved a 99.25% adjusted R2 score, highlighting the potential of machine learning tools in developing accurate and detailed fuel flow models based on full-flight data, providing significant benefits for the aviation industry.