Parameter optimization system for time series prediction

Abstract

The present invention belongs to the technical field of time series prediction, and particularly provides a parameter optimization system for time series prediction, in order to solve the technical problems that the prior art has high requirements on the prior knowledge, a relatively low expandable path, high time complexity, low practical feasibility and poor robustness. The parameter optimization system provided by the present invention comprises: a parameter optimization module, which is configured to perform parameter optimization on a pre-acquired time series prediction model based on a pre-built parameter optimization model. The parameter optimization module comprises a spatial control unit and a convergence control unit. The spatial control unit is configured to adjust a spatial search range of the parameter optimization module based on a first weight function; and the convergence control unit is configured to adjust a convergence rate of the parameter optimization module basedon a second weight function. According to the system provided by the present invention, distributed performance is added, individuals can communicate and collaborate efficiently, and the performance of the algorithm is improved.

Description

technical field [0001] The invention belongs to the technical field of time series prediction, and in particular relates to a parameter optimization system for time series prediction. Background technique [0002] With the development of technology, people will generate a large amount of data in their daily lives. Predicting and analyzing future trends based on historical data can greatly improve people's quality of life, so time series forecasting came into being. Time series forecasting is to predict and analyze future trends based on the time series of historical statistical data. Among them, the method of time series forecasting analysis is mainly through the establishment of a mathematical regression model. The parameters of the mathematical regression model are decisive for regression accuracy and generalization performance. Therefore, the quality of the parameters of the mathematical regression model is the key to determine the time series forecast. [0003] How to o...

AI Technical Summary

Problems solved by technology

[0003] How to optimize the parameters of the mathematical regression model is the key to determine the accuracy of time series forecasting. The existing parameter optimization methods mainly include direct determination method, cross-validation method, network search method and artificial bee colony algorithm, etc., but the existing parameter optimization methods are all There are some problems that are difficult to avoid. The direct determination method has high requirements for prior knowledge, and the expansion path is relatively low; the cross-validation method has high time complexity and low practical feasibility; the network search method consumes a lot of calculations, and it is difficult to guarantee convergence to The global optimal; the artificial bee colony algorithm is limited by its own evolution mode and selection strategy, the search accuracy is not high, the convergence speed is slow, and it is easy to be premature, etc. In addition, some traditional intelligent algorithms have high time complexity, poor robustness, and are easy to fall into local The optimal value

Images