Pharmaceutical industry, which has complex production processes and a serious VOCs emission problem, is a key industry for implementing VOCs emission reduction. To implement VOCs emission reduction, it is necessary to accurately identify key companies and production processes. Based on the VOCs source profiles of a typical chemical synthetic pharmaceutical factory in a fine chemical industrial park, the VOCs characteristic factors identification were carried out based on machine learning methods such as feature selection, classification analysis, and cluster analysis. The results showed that the VOCs emission characteristic factors were identified as toluene, acetone, acetaldehyde, benzaldehyde and n-hexane. The identified characteristic factors were obviously simplified by machine learning methods,and had similar concentration distribution in each production process. They could reflect the differences in the species composition of VOCs emission from pollution sources.