Multi-dimensional time sequence prediction system and method based on deep learning

• Main Authors: TAN FENG, FAN CHUNHAI, YU LONGXUAN, SONG HAITAO, LI JIAJIA, WANG ZIKAI
• Format: Patent
• Language: Chinese; English
• Published: 29.03.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention provides a multi-dimensional time sequence prediction system and method based on deep learning, the multi-dimensional time sequence prediction system based on deep learning comprises a data acquisition module, a multi-head jump loop network model training module and a prediction verification module, and the data acquisition module is used for acquiring multi-dimensional time sequence data; the multi-head jump loop network model training module comprises a one-dimensional convolution layer, a multi-head jump loop module, a multi-head jump convolution module, a fusion feature module and a prediction verification module, uses an autoregression layer to take original time sequence data as input, and fuses outputs of a dense layer and the autoregression layer to obtain a multi-head jump loop network model; according to the method, compared with a traditional autoregression prediction method, the medium-and-long-term prediction effect is more accurate, and compared with other light time sequence predi