Regression Forecast Model of 60-Meter Band Channel Performance Index Over Kupang Area

This paper describes the multivariate time series regression prediction model to forecast the performance of the 60-meter band channel over Kupang area. The frequency band which is used for medium range mode of the upcoming implementation NavDat communication system which extends the legacy of NavTex usability in maritime services by providing digital broadcasting of safety and security information. The hourly Near Vertical Incidence Skywave (NVIS) channel performance index over Kupang area was extracted from Ionosonde observation data. The regression models were implemented to predict 3-hour in advance of the NVIS channel performance. Together, in-situ data of ionospheric disturbance index or ionospheric scintillation index (S4) over Darwin, Australia with corresponding of Kp-index, Dst-index, the sunspot number and F10.7 solar flux data for the period of November 2021 to November 2022 were used as multivariate inputs for learning process. The results suggest that the Random Forrest regression model performed best compare to Extreme Gradient Boosting, Gradient Boosting, Decision Tree and Linear Regression model. The optimum model yields the following, mean square error (MSE) of 0.0011, root mean square error (RMSE) of 0.0326, mean absolute error (MAE) of 0.0238, and the coefficient of determination or prediction error (R2) value of 0.9725, respectively.