Experimental Comparison of Classical Machine Learning and Deep Learning Methods for RF-based Drone Classification

This paper presents an experimental comparison of classical machine learning (ML) and deep learning (DL) models for detecting and classifying drones based on radio frequency (RF) features. Although previous studies have explored individ-ual models or partial comparisons, inconsistent datasets have hindered direct performance evaluation between approaches. To enable a fair and comprehensive comparison, we collected a consistent RF dataset from three popular DJI drone models. Using this data set, we evaluated the performance of classical ML algorithms: Support Vector Machines (SVM), Random Forest, Gradient Boosting (XGBoost) and K-Nearest Neighbors (KNN); along with advanced DL architectures such as a shallow and deep Multi-Layer Perceptron (MLP). Our results demonstrate that DL models offer superior accuracy and robustness, whereas classical ML methods are more efficient in terms of training and inference time. This trade-off provides critical insight for selecting the appropriate models in practical drone detection and classification scenarios.