Machine Learning-Powered Satellite Trajectory Prediction: A Novel Approach for Enhanced Accuracy

Satellite trajectory prediction is essential for any space mission as it protects spacecraft from different collisions with natural celestials. Machine learning can plays a pivotal role in identifying the complex patterns of this prediction and improve the overall decision-making process. In this work, two different algorithms namely Random Forest and Deep ensemble model have been explored for satellite trajectory prediction. In Random forest model, study have been made with various number of decision trees and its impact on the overall performance of the model is studied. In Deep ensemble model, three LSTM regressors are used and finally voting regressor method is used. It is observed that both the models give almost comparable results. However, Deep ensemble model outperforms the Random forest result marginally with a $R^{2}$ value of 0.95. A comparative analysis of these models with non ensemble based models like linear regressor, decision tree regressor and SVM regressor showed that ensemble models perform better than non-ensemble based models