Method of computing confidence indicator for event state prediction and device of computing confidence indicator for event state prediction

Abstract

A method for calculating a confidence index for predicting the state of an event includes the following steps: repeatedly obtaining multiple first predicted values ​​from one of the prediction models for a specified time, and calculating single-model statistics for the multiple first predicted values; obtaining multiple second predicted values ​​from multiple prediction models for the specified time, and calculating multi-model statistics for the multiple second predicted values ​​from different models; inputting the single-model statistics into an index score converter to calculate a stability index; inputting the multi-model statistics into an index score converter to calculate a consistency index; and determining the confidence score of the multiple predicted values ​​for the specified time by using the stability index and consistency index associated with the specified time.

Claims

1. A method for calculating a credibility index for predicting event states, comprising: A processor repeatedly obtains multiple first predicted values ​​for a specified time from one of multiple prediction models, and calculates a first statistical value for the multiple first predicted values ​​as a single-model statistical value. The processor also obtains multiple second predicted values ​​for the specified time from each of the multiple prediction models, and calculates a second statistical value for the multiple second predicted values ​​from different models as a multi-model statistical value. Furthermore, the processor calculates a stability index, obtained by inputting the single-model statistical value into a linear or nonlinear converter of an index score converter. Finally, the processor calculates a consistency index, obtained by inputting the multi-model statistical value into the linear or nonlinear converter of the index score converter. Finally, the processor determines a confidence score between the multiple first predicted values ​​and the multiple second predicted values ​​for the specified time by correlating the stability index and the consistency index with the specified time.

2. The reliability index calculation method as described in claim 1, wherein the linear converter calculates the stability index by the following steps: calculating a first difference by subtracting the single-model statistical value from a maximum statistical value in a set of historical single-model statistical values; calculating a second difference by subtracting the minimum statistical value in the set of historical single-model statistical values ​​from the maximum statistical value in the set of historical single-model statistical values; and calculating a ratio between the first difference and the second difference, using the ratio as the stability index.

3. The reliability index calculation method as described in claim 1, wherein the linear converter calculates the consistency index by the following steps: calculating a first difference by subtracting the multi-model statistics from a maximum statistical value in a historical multi-model statistics set; calculating a second difference by subtracting a minimum statistical value from the maximum statistical value in the historical multi-model statistics set; and calculating a ratio between the first difference and the second difference, using the ratio as the consistency index.

4. The reliability index calculation method as described in claim 1, wherein the nonlinear converter calculates the stability index or the consistency index by the following steps: calculating a value of an exponential function based on the single-model statistical value or the multi-model statistical value and a sensitivity parameter, and using the value as the stability index or the consistency index.

5. The credibility index calculation method as described in claim 1, wherein the step of determining the credibility score by means of the stability index and the consistency index associated with the specified time includes determining that the credibility score is reliable when at least one of the stability index and the consistency index is greater than a threshold value; and determining that the credibility score is unstable when both the stability index and the consistency index are less than the threshold value.

6. The credibility index calculation method as described in Request 1, wherein the first statistical value and the second statistical value are a standard deviation or a variance.

7. A confidence index calculation apparatus for predicting event states, comprising: a transmission interface for receiving a plurality of predicted values ​​from a plurality of prediction models; a storage device coupled to the transmission interface and configured to store the plurality of predicted values; and a processor coupled to the storage device and the transmission interface and configured to perform the following operations: repeatedly obtaining a plurality of first predicted values ​​for a specified time from one of the plurality of prediction models, and calculating a first statistical value of the plurality of first predicted values ​​as a single-model statistical value; obtaining a plurality of second predicted values ​​for the specified time from each of the plurality of prediction models, and calculating a second statistical value of the plurality of second predicted values ​​from different models as a multi-model statistical value; and calculating a stability index obtained by inputting the single-model statistical value to a linear converter or a nonlinear converter of an index-score converter. Calculate a consistency index, which is obtained by inputting the multi-model statistics into the linear or nonlinear converter of the index score converter; and determine a confidence score between the plurality of first predicted values ​​and the plurality of second predicted values ​​at the specified time by using the stability index and the consistency index associated with the specified time.

8. The reliability index calculation apparatus as claimed in claim 7, wherein the linear converter calculates the stability index by means of the following steps: calculating a first difference by subtracting a single-model statistic from a maximum statistic in a set of historical single-model statistics; calculating a second difference by subtracting a minimum statistic from the maximum statistic in the set of historical single-model statistics; and calculating a ratio of the first difference to the second difference, using the ratio as the stability index.

9. The reliability index calculation apparatus as claimed in claim 7, wherein the linear converter calculates the consistency index by the following steps: calculating a first difference by subtracting the multi-model statistics from a maximum statistical value in a set of historical multi-model statistics; calculating a second difference by subtracting a minimum statistical value from the maximum statistical value in the set of historical multi-model statistics; and calculating a ratio of the first difference to the second difference, using the ratio as the consistency index.

10. The reliability index calculation device as described in claim 7, wherein the nonlinear converter calculates the stability index or the consistency index by means of the following steps: calculating a value of an exponential function based on the single-model statistical value or the multi-model statistical value and a sensitivity parameter, and using the value as the stability index or the consistency index.

Images