Um método para predição de taxa de transferência em Redes Móveis a partir de dados coletados de Smartphones

Abstract

Mobile communication in dynamic environments poses significant challenges to maintaining service quality, with data throughput being a critical performance parameter. Although throughput prediction in mobile networks has been widely explored for downlink, the uplink direction remains underexplored, despite the increasing demand from applications that require large volumes of data to be uploaded to the network. In this context, this work proposes a machine learning-based approach for predicting uplink throughput in 4G LTE networks, using physical layer metrics as predictive variables.A real-world urban data collection experiment was conducted, covering different times of day and traffic conditions, using an Android smartphone and the Armadeira application. The resulting dataset enabled the training and evaluation of five machine learning models, with the Random Forest algorithm achieving the best performance, reaching a coefficient of determination (R2) of 0.93 and the lowest mean absolute error.The results confirm the feasibility of the proposed approach, showing that the analysis of signal quality parameters, such as RSRP, RSRQ, RSSI, and SNR, enables accurate uplink throughput prediction. The adopted methodology supports the replicability of the experiment and provides a foundation for future research on mobile network optimization. This work advances technical and scientific knowledge on throughput prediction in mobile networks by providing a novel dataset, a validated method in a real environment, and practices that strengthen the analysis of Quality of Service (QoS) and Quality of Experience (QoE) in mobile communication systems.