In order to meet the communication needs of the new era, the new generation communication network should have the characteristics of faster speed, larger capacity and wider coverage. By mixing different modulation formats in communication networks, the requirements of new generation communication networks can be satisfied. Facing the situation that multiple modulation formats are mixed and variable in future communication networks, the correct identification of signal modulation formats can provide the basis for the subsequent use of digital technique to improve the performance. Machine learning has obvious advantages over traditional methods in feature extraction and classification. The application of machine learning techniques to realize modulation format identification has become a hot topic nowadays. This paper introduces several machine learning methods applied to feature extraction-based classification schemes, and try to summarize the field by analyzing and comparing their applications.