CELLULAR USER THROUGHPUT PREDICTION USING STATISTICAL AND MACHINE LEARNING MODELS

• Main Author: Syahri Ramadhani, Rifqi
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/65841
• Institution: Institut Teknologi Bandung

A survey conducted by Ericsson in November 2019 found that there are 5.9 billion mobile phone users and mobile traffic is expected to grow 27% annually from 2019 to 2025. This is highlighted by the upcoming Ericsson report for November 2021 which shows that mobile phone users will reach 6.7 billion users by the end of the year. This scenario has a causal effect on mobile traffic estimates reaching 65 EB/month and 11.4 GB/month per smartphone. Knowing the significant growth in network capacity mentioned above, it is necessary to adapt to the situation by making optimal use of mobile network capacity. The subject of optimization to be discussed is the downlink and uplink throughput of mobile users which fluctuates significantly over time. For observation, the observed physical objects are BTSs and their capacity to provide the required network capacity. Proper network capacity planning is essential to help network administrators evaluate the most cost-effective operations. Thus, this network planning can be facilitated using throughput prediction which will be the focus. During development, predictions are designed using statistical and machine learning models. However, since the number of BTS stations is unlimited, it is necessary to automate the calculation instead of doing the work manually one by one, so web development visualizes the dashboard to provide more calculations and a better planning experience. The details of the combined model will be decided on the basis of observing the model's performance through predictive objectives and measures for continuous data or regression problems.