Research on feature importance evaluation of wireless signal recognition based on decision tree algorithm in cognitive computing

Cognitive computing is an important method in the field of wireless signal processing, analysis and recognition. How to select features to complete the cognitive computing quickly and effectively is an important role in real application. In this paper, three kinds of features are extracted from six communication signals: power spectrum entropy, singular spectrum entropy and wavelet energy entropy. And the importance of the features is evaluated. Box-diagram and recognition rate are used for the evaluation of single feature. The visual boundaries of feature classification are used to evaluate two features. Meanwhile, the confusion matrix and the visualization model of decision tree are given for more detailed evaluation. The evaluation results show that the combination of power spectrum entropy and singular spectrum entropy can get the best recognition performance.