Abstract
This work introduces the use of multi-output regression algorithm for wave height and wave period prediction in the United States waters using the recorded data from 104 stations, from 2010 to 2019. The models use raw data for all the stations monitored by National Oceanic and Atmospheric Administration’s National Data Buoy Center. Five models are developed using four machine learning algorithms of K-Nearest Neighbors (KNN), Decision Tree (DT), Support Vector Regression (SVR), and Neural Network (NN). These models take a latitude, a longitude, and a month as inputs and predict three features, which are monthly maximum, monthly average and monthly minimum values for wave height and wave period at the given location in a given month. Results showed that the models developed based on DT, KNN, and NN algorithms have good performances, especially in terms of the monthly minimum and monthly average value prediction for both wave height and wave period values.