Oversampling based Classifiers for Categorization of Radar Returns from the Ionosphere
Surabhi Adhikari, Surendrabikram Thapa, Bickey Kumar Shah · 2020 International Conference on Electronics and Sustainable Communication Systems (ICESC) · 2020
Study of the ionosphere is important for research in various domains. Especially in communication systems, this study holds a great importance. In ionospheric research, there is a need to delineate useful and non-useful radar returns from the ionosphere. The useful radar returns can be used for further analysis and non-useful radar returns can be discarded. When the usefulness of radar returns is analyzed by humans, it is simply time-consuming and is prone to more human errors. Thus, some machine learning methods are needed to delineate useful and non-useful radar returns. The various machine learning algorithms (classical learning algorithms and ensemble learners) as well as deep learning models are tested in this study. After 10-fold cross validation, the highest accuracy by Random Forest was 94.22% and the accuracy for a single layer neural network was 99.43%.