The uncertainty quantification method of the data-based model using the apparatus for quantifying the uncertainty of the data-based model of the present invention calculates the verification data Euclidean distance (d v ) based on the memory data and the verification data, and uses the Monte-carlo method. model uncertainty of the verification data (x V) kernel weight (Kh v) and the verification data calculating a predicted value of uncertainty, and the verify data (x V) of the verification data (x V) calculating the model uncertainty and, by using the Gaussian kernel of the By comparing the uncertainty of the verification data prediction value, the optimization coefficient value can be optimized, the kernel weight (Kh q ) of the query data (x q ) by the query data Euclidean distance (d q ) is calculated, and the query data (x examples of q) the kernel weight (Kh q) and as compared to an optimized optimized coefficient values to obtain the number (N) of valid data, by the number (N) of the effective data of the data-driven model Calculating the uncertainty value, it is possible to calculate a highly reliable uncertainty than the previous model the uncertainty of the line can increase the reliability of the predicted value.
展开▼
