In order to explore a comprehensive evaluation method of food safety matching to the big data of supervised sampling test data， a fuzzy mathematics model for safety assessment of pesticide residue pollution of agricultural products was established based on the principle of entropy weight. In accordance with the large scale of sampling test data of pesticide residues in fruits and vegetables collected continuously from August 2017 to July 2019， an efficient comprehensive evaluation model was constructed and the evaluation effect of fuzzy mathematics model was studied by comparing with the traditional single-factor analysis. The results indicated that the overall safety of agricultural products in the market was at the level Ⅰ（micro risk）. Among them， the comprehensive risk of pesticide residue pollution of agricultural products was mainly determined by dimethylimide and benzimidazole pesticide in leafy vegetables （level Ⅱ， low risk） and bulbous vegetables （level Ⅱ， low risk）. This result was valuable for developing the supervision and sampling. Compared with the traditional analytical method， not only the comprehensive fuzzy evaluation method with the normalized and uniform evaluation system could achieve consistent evaluation results， but demonstrate better evaluation effectiveness for big data analysis. As a result， it is considerably applicable for the food safety evaluation and benefit to improving the efficiency of food safety supervision.