Epidemic situation prediction model based on LSTM (long short term memory) deep learning network model

Abstract

The invention discloses an epidemic situation prediction model based on an LSTM deep learning network model. The epidemic situation prediction model comprises LSTM network layers used for extracting rule information in sequence data, a full connection layer used for drawing up an output dimension, and an activation layer used for adjusting a fitting degree between prediction data and label data. The epidemic situation prediction model comprises two connected long short-term memory network layers, and the output of the first long short-term memory network layer is used as the input of the second long short-term memory network layer. According to the method, the 2* LSTM layer connection model is used for modeling serialized data, and the influence of changes between epidemic situation sequence data is fully take into consideration, so model fitting effect is better. After epidemic situation sequence data is subjected to training via one LSTM layer, memory information is reserved and transmitted to the next LSTM layer, so the model can fully learn change information between sequences, other parameters except epidemic situation change records do not need to be considered, prediction errors are relatively low, and the model has reference value for a later epidemic situation development trend.

Description

technical field [0001] The present invention relates to the field of deep learning algorithms, in particular to an epidemic prediction model based on an LSTM deep learning network model. Background technique [0002] With the prevention and control of the epidemic across the country, the epidemic is gradually under control. Predicting the development trend of the epidemic in the next few months will serve as a reference for whether the society will relax the prevention and control of the epidemic. Changes in epidemic data are closely related to government policies and the epidemic prevention and control situation of the masses, and there are certain fluctuations. The epidemic situation in various regions is roughly the same, but due to the relationship with the degree of supervision and the flow of people in the region, the data fluctuations are slightly different. [0003] At present, the most widely used SEIR (susceptible-exposed-infected-removed) model is used for epide...

AI Technical Summary

Problems solved by technology

The training results of the single-layer LSTM network model currently used in epidemic prediction are average, and there is a large difference between the predicted value and the real value

Images