Classification of oceanwave conditions using Support Vector Machine

Identifying ocean wave conditions is very important to marine related activities since it reflects the severity of current waves. One main issue is that for accurate identification of wave conditions, longer time series data are needed. However, these pose major impediment to real-time publishing of wave conditions. This study explores the possibility of classifying wave conditions given only shorter wave height time series data. By using shorter time series, systems similar to [1] will be able to handle the data faster and efficiently hence real time publishing of wave conditions can be achieved. The classification model for the wave conditions are trained using the Support Vector Machine (SVM) because it promises to provide a suitable model for nonlinear classification. Wave condition parameters are expected to have nonlinear relationships hence SVM is suited for this application. To test the ability of SVM, synthetic wave records generated from JONSWAP Wave Model are utilized. A classification map is then generated and compared to the ideal classification map of wave conditions.