Predicting wave parameters such as significant wave height (SWH), average wave period (APD) etc., is essential for a wide range of maritime, environmental, and economic applications. Reliable estimation of the parameters is needed to optimize the energy harvesting and protect infrastructures from damage. This paper attempts to predict and compare SWH and APD using machine learning (ML) based prediction models at NE Bahamas ( \(\mathbf {27.465}\) N, \(\mathbf {71.452}\) W) during the three years period 2021-2023. Additionally, the study attempts a novel approach to obtain the missing data for each month throughout the period, using MATLAB’s Piecewise Cubic Hermite Interpolation Polynomial(PCHIP) technique. The datasets considered in the present location are random in nature and possess significant challenges with a large number of missing data (day-wise, hour-wise) that show an irregular sampling interval which further adds to the error. In this study, SWH and APD is evaluated in two aspects: one by directly using the initial data in the ML models XGBoost, Random Forest, and Support Vector Machine, and two by updating the data by incorporating the missing data (interpolated data) using PCHIP and validating the results with the buoy measurements. Also, statistical error measures like RMSE, MAE, etc., were performed to show the improvement of the results with the inclusion of missing data in the ML model. This study enhances the performance of ML models by introducing the interpolated data using PCHIP in achieving better accuracy of the wave parameters in the presence of data with missing, irregularities, and extensive gaps.
