Application of seasonal ARIMA model in prediction of detection rate of norovirus in oyster

ObjectiveThe seasonal autoregressive integrated moving average (ARIMA) model was used to predict the detection rate of norovirus in oysters sold in Shanghai, which provided a reference for the prevalence of norovirus in aquatic products.MethodsOyster samples were regularly purchased from the Shanghai Luchaogang seafood market. A total of 531 oyster samples were tested for norovirus by nest-polymerase chain reaction (Nest-PCR), and the positive detection rate was calculated every quarter. The seasonal ARIMA model was used to fit the norovirus detection rate data in oysters from June 2016 to November 2019 to construct the model. After data stabilization, model selection and fitting and model diagnosis, the optimal model was obtained and the optimal model was used to predict the detection rate of norovirus in oysters in 2020.ResultsThe seasonal ARIMA (0, 1, 1) (0, 1, 0)4 was the optimal model. Akaike's information criterion and the finite corrections (AICc) (58.70) was the smallest. The residual error was a white noise sequence by Ljung-Box test. The trend of norovirus positive rate in oysters fitted by the model was basically consistent with the trend of actual detection rate, the mean absolute error (MAE) was 4.85 and the mean absolute percentage error (MAPE) was 30.25. The positive detection rates of norovirus in oysters predicted by the optimal model in the next four quarters were 31.89%, 12.80%, 9.47%, and 6.14%, respectively.ConclusionThe seasonal ARIMA model (0, 1, 1) (0, 1, 0)4 can fit the trend of positive detection rate of norovirus in oysters. This model has certain significance for the risk assessment of aquatic products such as oysters contaminated by norovirus. It also has certain significance for the prevention and control of the norovirus epidemic.