Title : Low-dimensional analysis of environmental factors mediating norovirus outbreaks from food
Abstract:
Introduction
Contaminated oysters are a severe threat to public health in the US, with Norovirus outbreaks being a common concern. To prevent these outbreaks, predicting when they might occur is critical. Fortunately, advanced artificial intelligence models can now forecast contaminated oyster outbreaks up to two weeks in advance. These models account for six environmental factors - water temperature, salinity, solar radiation, rainfall, wind, and gage height. To enhance the accuracy of these predictions, our team is currently conducting a reduced-order manifold, traditionally known as POD or principal component analysis (PCA) on meteorological data. It may be possible to reduce the required input variables for accurate predictions by training an artificial neural network (ANN) using PCs. The potential of such an approach is discussed in the present work.
Methods
Quantitative secondary data from the National Outbreak Reporting System and other state and federal agencies for more than 2 decades are collected from the most popular Oyster harvesting area around the Gulf of Mexico. Pre-processing of the data composed of data cleaning and statistical analysis, including dimensionality reduction, is conducted. ANNs are trained towards the precise prediction of a fraction of the preceding outbreaks in the areas of interest. The training step is followed by predicting the outbreaks from the data that has yet to be presented to the ANN during the training phase.
Results
Based on the preliminary data analysis, it is evident that the proposed method possesses enormous potential in systematically selecting meteorological data for future forecasts.
Conclusion
In summary, training an artificial neural network with principal components could lead to an ANN model capable of forecasting norovirus outbreaks caused by contaminated oysters with fewer environmental predictors than current models. This could allow for predicting norovirus outbreaks in areas with limited meteorological data.
