Predicting nearshore wave direction plays a vital role in coastal management, maritime operations, and environmental monitoring. The study explored a machine learning approach to predict wave direction using two popular ensemble models: The Random Forest Regressor and Gradient Boosting. The data used for training includes wave height, period, and wind speed which are key wave parameters, and all of which were carefully preprocessed. The Study assessed the performance of both models using two main metrics: Root Mean Squared Error (RMSE) and R-squared (R2). The Gradient Boosting model yielded an RMSE of 7.81 and R2 score of 0.76. The Random Forest model performed less, achieving an RMSE of 8.27 and an R2 score of 0.72, with a slight improvement in accuracy. To better understand the models’ behavior, we employed visualization tools like residual plots, which helped identify where errors occurred, feature importance charts to highlight the most influential factors with feature 0 as Significant Wave Height, features 1 as Wave Periods and feature 2 as Wave Up-crossing Period, and wave direction plots to compare the actual and predicted values. Interestingly, the feature analysis revealed that wave height was the strongest predictor in both models. Overall, by combining Random Forest and Gradient Boosting, this study offers a reliable and interpretable approach for predicting wave direction, with promising applications in coastal engineering and other real-world scenarios.