Multi-scale time series prediction method based on adaptive sparse expert selection strategy and closed continuous time neural network

By combining adaptive sparse expert selection and closed continuous-time neural networks, the problems of single time scale and sparse expert routing in multivariate time series forecasting are solved, realizing multi-scale adaptive modeling and global continuous-time modeling, thus improving the efficiency and accuracy of forecasting.

CN121542585APending Publication Date: 2026-02-17HUNAN UNIV
0 Cites 1 Cited by

Patent Information

Application Number
CN202610067790.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-01-19
Publication Date
2026-02-17

AI Technical Summary

Technical Problem

Existing multivariate time series forecasting methods suffer from limitations in handling single time scales, the impact of sparse expert routing mechanisms on the continuity of time series modeling, the difficulty in unifying the modeling of local and global features, the lack of adaptive time scale modeling capabilities, and the absence of efficient sparse expert selection mechanisms. Consequently, these methods exhibit insufficient adaptability and generalization performance in complex dynamic systems.

Method used

By employing an adaptive sparse expert selection strategy and a closed continuous-time neural network, we achieve multi-timescale adaptive modeling and global continuous-time modeling through RevIN normalization and low-dimensional feature mapping, multi-scale trend-seasonal structure enhancement, local time feature modeling of multi-scale expert models, adaptive sparse expert selection and load balancing loss calculation, weighted aggregation and residual fusion of multi-scale expert outputs, and global modeling based on channel-weighted closed continuous-time neural networks.

Benefits of technology

It achieves efficient and stable multivariate time series prediction, improves the model's ability to model complex time series structures, enhances the capture of long-term dependencies and the representation of global features, and improves the model's adaptability and generalization performance in complex dynamic systems.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a multi-scale time sequence prediction method based on an adaptive sparse expert selection strategy and a closed continuous time neural network. The method comprises the following steps: carrying out normalization and low-dimensional feature mapping based on RevIN; performing trend-seasonal structure enhancement processing on the feature sequence after linear mapping; constructing a multi-scale expert model based on the feature sequence after trend-season enhancement; self-adaptive sparse expert selection and load balancing loss calculation are carried out; carrying out weighted aggregation and residual fusion on multi-scale expert output; global modeling of a closed continuous time neural network based on channel weighting is carried out; and finally performing prediction generation and reverse normalization. The multi-scale time series prediction method has the multi-time-scale adaptive modeling capability, the sparse expert efficient selection mechanism and the global continuous time modeling capability, and can be applied to various multivariable time series prediction scenes such as power load prediction, weather prediction, industrial production monitoring, traffic flow prediction and financial price prediction.
Need to check novelty before this filing date? Find Prior Art