Ship fuel efficiency and CO2 emissions analysis plays a vital role in enhancing thesustainability of maritime transportation and mitigating its environmental footprint. However, the traditional methods have difficulty in adapting real time data and the inabilityto capture complex and non-linear data because of relying on simplified assumptions and historical data. Fuel efficiency and CO₂ emissions are influenced by complex and nonlinear interactions between variables like weather conditions, engine performance, vessel load, andsea conditions. These traditional methods cannot easily model such intricate relationships, leading to simplified approximations that may not fully capture real-world dynamics. Deep learning (DL)-based methods could somewhat identify subtle, nonlinear dependencies between factors that influence fuel consumption and CO₂ emissions, leading to more accurate and reliable predictions. However, these models face limitations in training the noisy, sparse and turbulent data patterns, which reduces the accuracy of predicting the output and the computation time is increased. Therefore, this paper presents an Artificial Neural Networks using Adaptive Moment Estimation (ADAM) optimization algorithm to overcome the drawbacks faced due to noisy and sparse dataset. The ADAM optimizer enhances the learning process by adjusting learning rates based on first and second moments of gradients, offering faster convergence and improved resilience to noisy and sparse data. Using an ANN optimized with ADAM algorithm showed an improvement in prediction efficiency by 8% to
15% compared to Deep Learning Models