[Application of Best Subsets Regression on the risk classification for Spermophilus Dauricus Focus]

To study the risk classification of animal plague in Spermophilus Dauricus Focus, using the Best Subsets Regression (BSR) model.

Matlab, BSR and exponential smoothing were employed to develop and evaluate a model for risk classification as well as to forecast plague epidemics at the Spermophilus Dauricus Focus. Data was based upon the Inner Mongolia surveillance programs. This model involved 7 risk factors, including density of Spermophilus dauricus, percentage of hosts infested, host flea index, percentage of nests infested, nest flea index, percentage of runways infested, and runway flea index.

Forecasting values of the classification model(CM)were calculated and grouped into 3 risk levels. Values that over 2/3 of the CM would indicate the existence of potential epidemics while those below 1/3 would indicate that there were no risk for epidemics but when values that were in between would indicate that there exist for high risk. Annually, during the observation period in the Inner Mongolia Spermophilus Dauricus Foci, the detection of Yersinia pestis gave a risk rating value of 1 which stood for existing epidemics, while nil detection rate generated a 'zero' value which representing the situation of non-epidemic. The overall plague epidemics forecasting surveillance programs in 2012 at the Spermophilus Dauricus Foci indicated that no active plague was observed. When the forecasting values became over 2/3, combinations of all the risk factors would achieve the consistency rates of 100%. When the forecasting values were below 1/3, combinations of at least the first 4 factors could also achieve the consistency rates of 100%. However, when the forecasting values fell in between, combinations of at least the first 4 factors would achieve the consistency rates of around 50%.

Results from our study showed that plague would not be active to become epidemic, in 2012.