5G NETWORK TRAFFIC FORECASTING USING MACHINE LEARNING

Abstract

The idea of network chunks being described as virtual subsets of the physical resources of 5G infrastructure is used in standards for 5G communications. The efficiency of ML predictors for traffic prediction in 5G networks has been established in recent research so that it becomes to assess the capability demands of each network slice and to see how it progresses as a large number of network slices are deployed over a 5G network over time to be very important. The main objective of this research is to establish the model that has the potential to help network management and resource allocation in 5G networks with machine learning performance analysis in predicting network traffic on high-dimensional spatial-temporal cellular data, in addition to investigating the effectiveness of various neural network models in traffic prediction from univariate and multivariate perspectives. The research method used is a quantitative research method using correlation analysis, statistical analysis, and distribution analysis on the temporal and spatiotemporal frameworks developed to predict traffic from a univariate and multivariate perspective. To predict 24-hour mobile traffic requires combining spatial and temporal dependencies. The univariate analysis will be carried out by applying a temporal framework that includes FCSN, 1DCNN, SSLSTM and ARLSTM to capture temporal dependencies. The results of various experiments in this study show that the proposed spatiotemporal model outperforms the temporal model and other techniques in the mobile traffic forecasting literature including internet, SMS, and calls.