Abstract:
Food safety inspection data contains rich information, which plays a great contribution in early warning and risk prediction of food safety. To deeply study the high-weight adulterants in food categories, we designed a classification problem based on contrast learning based on feature selection techniques, combined with correlation metrics between features, and finally implemented an adulterant correlation analysis visualization system. The system is designed and implemented with multiple views and linkage interactions for sample classification results, adulterant feature weights, adulterant feature correlations and model evaluation metrics, and can support users to get the optimal feature combinations based on the above information in an iterative analysis. Finally, we combine a real food safety testing dataset to demonstrate that the system can be automated to obtain adulterant weights of food products more easily and directly, and enhance the adulterant feature combinations.