Time series analysis of human and bovine brucellosis in South Korea from 2005 to 2010

Human incidence rate was highly correlated to cattle incidence rate in the same month and the previous month (both r = 0.82). In the final models, ARIMA (0, 1, 1) − AR (0, 1) was determined as the best fit with 191.5% error in the validation phase, whereas the best NBR model including lags (0, 1 months) for the cattle incidence rate yielded a 131.9% error in the validation phase. Error (MAPE) rates were high due to small absolute human case numbers (typically less than 10 per month in the validation phase). The NBR model however was able to demonstrate a marked reduction in human case immediately following a hypothetical marked reduction in cattle cases, and may be better for public health decision making.