Abstract
Spectrum occupancy prediction is a key enabling technology to facilitate a proactive resource allocation for dynamic spectrum management systems. This work focuses on the prediction of duty cycle (DC) metric that reflects spectrum usage (in the time domain). The spectrum usage is typically measured on a shorter time scale than needed for prediction. Hence, data thinning is required and we apply block averaging. However, averaging operation results in flattening the DC data and losing essential features to assist deep neural network (DNN) to predict the spectrum usage. To improve DC prediction after block averaging, a feature-based deep learning framework is proposed. Namely, long short-term memory (LSTM) and gated recurrent unit (GRU) are selected and enhanced by using features of the data, such as the variance of DC data in addition to DC data themself. The proposed model is capable of proactively predicting the spectrum usage by capturing complex relationships among various input features for the measured spectrum, thus providing higher prediction accuracy with an average improvement of 5% in RMSE compared with traditional models. Moreover, to have a better understanding of the proposed model, we quantify the effect of input features on the predicted spectrum usage values. Based on the most significant input features, a simpler and more efficient model is proposed to estimate DC with similar accuracy to when using all features.
Highlights
Unlike previous generations in wireless communications, the next-generation wireless networks are expected to be fast and capable of connecting several billions of devices
In 5G, three distinct use cases are defined by 3rd Generation Partnership (3GPP), namely enhanced Mobile Broadband, Ultra-Reliable and Low Latency Communications (URLLC), and massive Machine Type Communication
long short-term memory (LSTM) The LSTM network is a variation of recurrent neural networks (RNNs) which is typically used for time series data types
Summary
Unlike previous generations in wireless communications, the next-generation wireless networks are expected to be fast and capable of connecting several billions of devices. For the time series prediction problem, several studies concluded that using only traffic information as input feature is not sufficient, in [25] it was shown that using only traffic as input feature for enterprise network traffic prediction did not provide any advantage over traditional linear regression methods such as autoregressive integrated moving average (ARIMA). It can be concluded that having more input features to the neural network such as spatial details of transmitters and users, number of active users, type of device used for internet or network access and other statistical features such as the ones described in [29] will impact the predicted traffic usage.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have