Objective: Solve the problems of low and inaccurate evaluation accuracy of the existing food safety risk evaluation methods. Methods: Based on the construction of food safety risk index, a food risk assessment method combining hidden Markov model and cuckoo search algorithm was proposed. The cuckoo algorithm was used to search the global optimal solution as the initial value of the hidden Markov model, and the Baum Welch algorithm was used for local correction to make it quickly converge to the global optimal solution. The superiority of this method was verified by the experimental analysis of dairy product data. Results: Compared with that before the improvement, the improved evaluation method was more accurate and effective, which could more accurately predict food quality and safety under the time characteristics, and the detection accuracy was more than 97.5%. Conclusion: This method combines time characteristics to predict, has high detection accuracy, and can provide effective risk assessment results for enterprises.