For most rice, it is based on the classification of pictures alone, and it is not suitable for the evaluation of rice products on e-commerce website platform. A rice evaluation method based on cross-media is proposed. The two sets of feature vectors extracted from rice images and rice reviews were used to study the fusion classification. Firstly, the outline features of rice images and the feature word vectors of rice reviews were extracted, and then the two sets of vectors were merged by using typical correlation analysis. The merged features are standardized, and the random forest is used as a classifier to learn. Finally, a rice evaluation method based on cross-media multi-feature fusion is realized. By learning and testing the rice on the e-commerce website, the proposed method achieves 90% accuracy, compared to the random forest under multi-features and the KNN under the multi-features, and the tree classifier. High accuracy.
展开▼