Comparative Analysis of Machine Learning Algorithms in 5G Coverage Prediction

• Author(s): Varsha N ; Suhas K C ; Harish T A
• Paper ID: 1710183
• Page: 626-634
• Published Date: 22-08-2025
• Published In: Iconic Research And Engineering Journals
• Publisher: IRE Journals
• e-ISSN: 2456-8880
• Volume/Issue: Volume 9 Issue 2 August-2025

Abstract

The development of 5G technology is the next rad- ical change in the sphere of telecommunication, which promises to virtually eliminate the speed of data transmission, provide an insignificant level of latency, and enable a very large quantity of devices to be connected at once. Nevertheless, it is a challenge to provide the best and reliable 5G signal coverage as the network gets more complex with dense infrastructure. Conventional ways of planning networks and validating coverage through methods like field testing and making use of propagation modeling are not only time consuming, highly expensive but also unable to respond to real time changes in the environment. In this project, the idea is to present a machine learning approach to intelligent, automated, and predictive measurement of 5G signal strength. The idea is to create the system which is able to predict signal quality of network based on such parameters as Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ), Received Signal Strength Indicator (RSSI), downlink and uplink bitrates, CELLHEX, NODEHEX, and geographic coordinates (latitude and longitude). Depending on these parameters, the signal strength is categorized into three, namely: Strong, Moderate, and Poor coverage. In order to come up with this we perform a comparative analysis of a range of supervised machine learning models, i.e. Random Forest, Support Vector Machine (SVM), Logistic Regression, AdaBoost, K- Nearest Neighbours (KNN), Gaussian Naive Bayes, LightGBM and ensemble models such as Voting Classifier and Stacking Classifier. These models are compared on the basis of their prediction accuracy with appropriate real time implementation. In order to produce a better model performance and reliability the dataset is fully preprocessed. Among them are treatment of the missing values, outlier deletion with the help of Interquartile Range (IQR), a coding of categorical features, and Synthetic Minority Oversampling Technique (SMOTE) to correct the issue of class imbalance. This is a smart solution that seeks to assist telecom operators in real-time and precise large-scale 5G network design and development optimization.

Citations

IRE Journals:
Varsha N , Suhas K C , Harish T A "Comparative Analysis of Machine Learning Algorithms in 5G Coverage Prediction" Iconic Research And Engineering Journals Volume 9 Issue 2 2025 Page 626-634

IEEE:
Varsha N , Suhas K C , Harish T A "Comparative Analysis of Machine Learning Algorithms in 5G Coverage Prediction" Iconic Research And Engineering Journals, 9(2)