Ocean Forecasting with Machine Learning Course

This short course combines data and artificial intelligence for ocean forecasting. You will distinguish physical, statistical and machine learning approaches and forecast horizons, prepare buoy and satellite data with sliding windows, feature engineering, temporal splits, normalisation and missing values, and train linear regression with regularisation, decision trees, random forests and gradient boosting with hyperparameter tuning and feature importance. You will then apply dense, recurrent, LSTM and convolutional neural networks, correct numerical model output, produce probabilistic forecasts and validate them with error and skill metrics against reference forecasts, including extreme events.

What sets this course apart: a practical approach with models trained on open ocean data; an honest comparison of results against reference physical forecasts; rigorous validation with metrics and schemes suited to time series; open tools such as Python, scikit-learn and deep learning libraries; and a practical case in which you build a forecasting model for a real variable and location.