Interval type index forecasting method based on robust interval extreme learning machine

Abstract

The invention discloses an interval type index forecasting method based on a robust interval extreme learning machine, belongs to the fields of automatic control, information technologies and advanced manufacturing, and particularly provides an interval type index forecasting method for achieving production index forecasting under the conditions that as for the production process in the uncertain environment, data comprise abnormal points, and indexes can be described through the number of intervals. The interval type index forecasting method is characterized by including the following steps: an interval upper-bound model and an interval lower-bound model are built through the interval extreme learning machine, and an optimization problem is built to optimize parameters of the two models, wherein an objective function of the optimization problem considers the model complexity, the model center error and the model interval error. In order to reduce influences of the abnormal points on the index forecasting performance, the abnormal points in an initial training data set is processed with a least median square method in robust statistics to determine sub-training-data-sets. The interval type index forecasting method can be used for forecasting the interval type production indexes with the abnormal points contained in the data.

Description

technical field [0001] The invention belongs to the fields of automatic control, information technology and advanced manufacturing, and specifically relates to the problem of index forecasting for the production process in an uncertain environment, where the data contains abnormal points and the production index can be described by interval numbers, and proposes a method based on robust interval limit learning Machine interval index forecasting method. Background technique [0002] In the operation optimization and dynamic scheduling of the actual complex production process in steel, microelectronics and other industries, it is often necessary to forecast the production index. However, due to the large uncertainty in the actual production process and the production data often contain abnormal points, the method based on neural network, There is often a large deviation between the index forecast value given by support vector machine and other conventional forecasting models a...

AI Technical Summary

Problems solved by technology

[0002] In the operation optimization and dynamic scheduling of the actual complex production process in steel, microelectronics and other industries, it is often necessary to forecast the production index. However, due to the large uncertainty in the actual production process and the production data often contain abnormal points, the method based on neural network, There is often a large deviation between the index forecast value given by support vector machine and other conventional forecasting models and the actual measurement value of the index, which affects the operation optimization and dynamic scheduling effect. The use of interval index forecasting method is an effective way to solve the above index forecasting problems one

Images