Click farming has become a common phenomenon, which brings great harm to the online shopping platform and consumers. To identify click farming on the Taobao platform, the largest online shopping platform in China, we use the positive-unlabeled learning method to find reliable negative instances from the unlabeled set and output the identification of click farming with probability rank for all shops, after creating several features from both goods and online shops. Then, a weighted logit model is used to investigate the role of extracted features in dissecting click farming. The empirical findings show that the extracted features are efficient to identify and explain click farming. And, the results show that click farming may not necessarily depend on the state of the shop. Our study can help online consumers to reduce the risk of being deceived, and help the platform to improve its regulatory capacity in click farming.