Deep learning model and flow data prediction method thereof

The embodiment of the invention provides a deep learning model and a flow data prediction method thereof. The deep learning model may include an input layer, an encoder, a decoder, and a prediction layer. N original feature sets corresponding to N time periods arranged according to a time sequence and N-1 flow data corresponding to the first N-1 time periods can be acquired through an input layer,wherein the single original feature set comprises M attribute features influencing the traffic data. Then, through an encoder comprising an encoding attention layer and a recursive encoding layer, anattention mechanism is added in the input stage of the encoder, the N original feature sets are corrected in a self-adaptive mode to obtain N corrected feature sets, and recursive processing is conducted on the N corrected feature sets to obtain N state vectors; and recursive processing is carried out on the first N-1 state vectors and the N-1 flow data by a decoder to obtain a target decoding vector. And finally, the target decoding vector and the Nth state vector are processed through a prediction layer to obtain flow data corresponding to the Nth time period.