Base Station Traffic Prediction Using Wavelet Transform and Bi-LSTM

Global mobile traffic is growing exponentially with the rapid deployment of 5G networks. Predicting traffic can better allocate the number of base station carrier frequencies, so as to alleviate the pressure on base stations due to data growth. The base station traffic data can be abstracted as time series with the characteristics of trend, periodicity, and randomness, which makes it suitable for predicting by the deep learning models, especially the Bi-LSTM. But the Bi-LSTM needs to rely on a large number of training sets to consume more resources, and are susceptible to outliers. Therefore, to improve the learning efficiency and prediction accuracy, this paper presents a new method that first uses decomposition and reconstruction of wavelet transform to preprocess the data set and then uses the Bi-LSTM model to predict the traffic. The experimental results show that the method has good feasibility and superiority in predicting base station traffic.