Intelligent selection method for ensemble prediction

Abstract

The present invention provides an intelligent selection method for ensemble prediction. The ensemble prediction provides a plurality of predictions of a certain numerical value, and the plurality of predictions form a one-dimensional random vector, so that an actual value is predicted as accurately as possible. The intelligent selection method for ensemble prediction comprises: performing classification statistics on historical data; then storing classification statistical information; and assessing the accuracy of ensemble prediction by comparing the statistical information, thereby selecting an accurate prediction from the predictions. According to the intelligent selection method for ensemble prediction, on the premise of disorderly ensemble prediction data, the ensemble prediction data are processed to select prediction data with the possibly smallest error with the actual value.

Description

technical field [0001] The present invention generally relates to the fields of data statistics and data mining processing, and specifically relates to an intelligent selection method for set prediction. Background technique [0002] Prediction is a scientific investigation and analysis of the history and current situation of objective facts, inferring the future from the past and the present, and inferring the unknown from the known, thereby revealing the trends and laws of the future development of objective facts. We cannot wait until something happens and then act without predicting, so good forecasting is very necessary. However, when it comes to forecasting, no single forecasting method is absolutely valid. Regardless of the method used for forecasting, forecasting is limited and not perfect. [0003] The ensemble prediction method overcomes the shortcomings of a single prediction and provides multiple prediction data for the prediction of real data, but sometimes th...

AI Technical Summary

Problems solved by technology

[0003] The ensemble prediction method overcomes the shortcomings of a single prediction and provides multiple prediction data for the prediction of real data, but sometimes the multiple prediction values ​​given by the ensemble prediction are very different, and some prediction values ​​deviate greatly from the actual value. Need a good way to judge which of these data is closer to the actual value

Images