Wireless spectrum occupancy prediction method based on LSTM network

Abstract

The invention relates to the field of wireless spectrum analysis, in particular to a wireless spectrum occupancy prediction method based on an LSTM network. According to the method, the ARIMA model and the Kalman filtering algorithm are combined to overcome the limitation of the ARIMA model, and the initial value of the Kalman filtering algorithm is determined by the ARIMA model and complements each other. Considering that the LSTM neural network has a very strong capturing capability on the nonlinear relationship,, an ARIMA, Kalman and LSTM combined prediction model is constructed, namely, alinear relationship existing in frequency band occupancy rate sequence data is extracted by utilizing an ARIMA and Kalman hybrid model, an unextracted nonlinear part in residual errors of the hybrid model is extracted by LSTM, and fitting information is superposed into the ARIMA and Kalman hybrid model. According to the method, the advantages of ARIMA, Kalman and LSTM networks are combined, the stationary sequence can be analyzed, the non-stationary sequence can also be analyzed, and meanwhile linear and nonlinear information can be well extracted.

Description

technical field [0001] The invention relates to the field of wireless spectrum analysis, in particular to an LSTM network-based wireless spectrum occupancy prediction method, which uses LSTM to predict wireless spectrum occupancy, and uses ARIMA combined with Kalman's hybrid model to improve it. Background technique [0002] At present, there are many methods for application and spectrum prediction, which can be summarized into two categories: [0003] 1) Based on statistical analysis methods: In 2012, Wang Lei et al. aimed at the problem that the traditional spectrum occupancy autoregressive moving average (ARMA) model could not accurately describe the nonlinear time-varying characteristics of the spectrum occupancy state because it did not consider the conditional second-order moment of the sequence. This paper proposes a time series modeling method of spectrum occupancy state based on Exponential Generalized Autoregressive Conditional Heteroscedasticity (EGARCH) process. ...

AI Technical Summary

Problems solved by technology

[0006] In view of the above-mentioned problems or deficiencies, in order to solve the relatively single and applicability problems of existing wireless spectrum occupancy prediction methods, the present invention provides a wireless spectrum occupancy prediction method based on LSTM network, by fully combining traditional spectrum occupancy prediction The advantages of the method and neural network can effectively realize the prediction of spectrum occupancy, and can take into account the extraction of linear information and nonlinear information and the processing of non-stationary sequences

Images