Traffic Prediction Based on Random Forest Regressor for Efficient Resource Allocation in Cellular Networks

In recent years, the quick development of cellular technologies such as 5G and Internet of Things (IoT), has led to a rapid increase in cellular network traffic. However, the existing Support Vector Regressor (SVR) faced challenges in task offloading time forecasting and modelling for increasing the Quality of Service (QoS). To overcome these challenges, the research proposes Machine Learning (ML)-based Traffic Prediction (TP) model with Random Forest Regressor (RFR) by learning by handling complex non-linear relationships effectively to predict load traffic in cellular networks. Initially, the data is collected from In-phase and Quadrature (IQ) samples dataset which consists of signal data based on bandwidth-related characteristics and load conditions of network traffic. Further, the data is preprocessed with High-Pass (HP) and Low-Pass (LP) filters by removing high-frequency noise and reducing low-frequency vibration. After that, Power Spectral Density (PSD) is utilized to measure the distribution across various frequencies by converting time domain into frequency domain. The proposed Random Forest Regressor (RFR) is utilized as predictive model, then Proportional Bandwidth Allocation (PBA) is used for resource allocation based on predicted offloading time. From the results, the proposed RFR model achieved outstanding results in terms of R^2 (99.26), RMSE (0.346), and MAPE (0.025) when compared with the existing SVR.