Supervised Abnormal Signal Identification Method

Identifying radio anomalies is one of the main purposes of radio monitoring. The current radio anomaly signal identification is mainly finished manually by the Radio monitor, using professional radio knowledge and their work experience. However, because the anomaly signal is hidden in the "massive" data, accompanied by a large amount of noise, and also data imbalance, the anomaly signal is difficult to find. In this paper, we combine the data unevenness processing method SMOTE and the support vector machine (SVM), gradient lifting tree (GDBT), and other classification algorithms to identification the anomaly signal during radio monitoring. Experimental results show that our method can improve the efficiency of existing radio anomaly signal recognization. Moreover, our experiments also show that data imbalance processing plays a key role in anomaly signal recognition.