Data Balancing in SAT-NET: Effects on Logistic Regression and Random Forest Performance

In the context of Satellite Internet Services and Applications (SAT-NET), data imbalance presents a significant challenge, severely impacting the performance of classification algorithms. This study investigates the impact of data balancing techniques, particularly the Synthetic Minority Over-sampling Technique (SMOTE), on the classification accuracy of logistic regression and random forest algorithms. Experimental results reveal that while SMOTE significantly reduces the accuracy of the logistic regression algorithm, it has minimal impact on the random forest algorithm, which maintains stable performance on both balanced and imbalanced datasets. This research provides empirical insights into the role of data balancing techniques in different classification algorithms and offers valuable guidance for algorithm selection and data balancing methods in SAT-NET data processing. Future research directions include exploring combinations of various data balancing techniques and their effects in different SAT-NET application scenarios to further enhance classifier performance. Keywords- Satellite Internet data imbalance; SMOTE; logistic regression; random forest; classification accuracy.