Enhancing Ionospheric Radar Returns Classification with Feature Engineering-Based Light Gradient Boosting Machine Algorithm

In ionospheric research, the distinction between relevant and irrelevant radar echoes is pivotal. Utilizing beneficial echoes for further analysis while disregarding irrelevant ones. Traditional manual analysis methods, being time-consuming and error-prone, have paved the way for the adoption of machine learning methods as a more efficient alternative. This paper introduces a Feature Engineering-based Light Gradient Boosting Machine (LGBM) algorithm designed for the classification of ionospheric radar echoes. The incorporation of advanced features, unearthed through feature engineering, brings a novel interpretive perspective and methodology to comprehend the dataset. The classification process employs the LGBM algorithm. The results obtained through this approach showcase a remarkable accuracy of 97.30% and an F1-score of 97.7%, highlighting outstanding performance in ionospheric radar returns classification.