Abstarct: Abstract Coastal areas hold significant economic, social, and environmental value. However, they are also highly vulnerable to threats such as erosion, storms, and rising sea levels. This study aims to analyze coastal morphodynamic changes, with a particular focus on berm and shoreline variations under non-storm conditions, using machine learning algorithms. Examining changes in the coastal berm crest under non-storm conditions is essential, as these conditions dominate most of the year and directly influence long-term erosion trends and coastal stability. Moreover, understanding these variations contributes to the effective management of coastal areas, the monitoring and mitigation of adverse human impacts, and the enhancement of predictive models for storm conditions. Field data on morphodynamics and hydrodynamics from the Narrabeen Coastal Station in Australia have been utilized to achieve the research objectives. The study period spans 2006 to 2019 and includes primary data on coastal profiles, wave characteristics, and sea level variations. Subsequently, the key factors influencing the research objectives have been analyzed and assessed to integrate these data into the modeling process. In this regard, machine learning algorithms—encompassing feature selection, classification, and prediction algorithms—have been employed as the primary tools in the modeling frxamework. To reduce model dimensions and eliminate irrelevant features, the Chi2 feature selection algorithm was employed. Subsequently, the Support Vector Machine (SVM) algorithm was used to classify the significant features. Predictive models were developed using Random Forest (RF), XGBoost, and Multilxayer Perceptron (MLP) algorithms. Results showed that the RF algorithm outperformed others in predicting all targets. For shoreline changes, Scenario 4, which included features like wave power, berm width changes, maximum wave height, and wave energy, achieved an RMSE of 3.32 meters and an R² of 88%. In predicting berm crest elevation changes, Scenario 4, incorporating features such as shoreline changes, berm crest horizontal position, wave power, and beach slope, resulted in an RMSE of 0.22 meters and an R² of 74.8%. Additionally, for predicting the berm crest horizontal position, Scenario 5, which included features like berm width, sea level rise, berm crest elevation, wave breaking index, and beach slope, provided the best performance with an RMSE of 8.37 meters and an R² of 80.1%. This study highlights the application of machine learning-baxsed methods for monitoring coastal changes under non-storm conditions, contributing to a better understanding of coastal dynamics. The results demonstrate the effectiveness of features classification in enhancing predictive algorithm performance and emphasize the importance of morphodynamic features in these analyses. Moreover, the Random Forest algorithm is identified as a robust and reliable tool for coastal predictions.