In order to effectively identify whether the National Geographic Indication rice（hereinafter referred to as the landmark rice） is mixed with ordinary rice, three machine learning identification models including SVM, AdaBoost and AdaBoost-SVM were established based on the intermediate fusion data of mineral element content and near infrared spectrum. The result shows that all the three models have excellent discriminating ability. The SVM model is better than the other two models in small proportion（2%-6%） discrimination, and the accuracy of the 100%. Adaboost model is better than that of the other two models in the selection of optimal fusion data sets. The lowest detection rate of the three models can be up to 2%, and the accuracy of the three models is 100%, 100% and 97.75%, respectively. Data fusion technology combined with machine learning method can be used as a reliable tool for accurate identification of rice adulteration, and provide technical support for maintaining the healthy and orderly development of rice market.