Skip to Main content Skip to Navigation

Spatiotemporal Individual Mobile Data Traffic Prediction

Abstract : Understanding the nature of data network traffic is critical in network design, man- agement, control, and optimization. In this report, we leverage two large-scale real-world datasets collected by a major mobile carrier in a Latin American country to investigate the prediction of individual mobile data traffic. Based on our previous analysis on the theoretical predictability, we extend our analysis to the actual prediction and validate the findings, that we have observed in the theoretical analysis, in the actual predicting scenario. We implement the typical algorithms for time series prediction in the literature and test their performance. Then, we propose our algorithms based on state-of-the-art machine learning techniques. Our data-driven test on the performance of these predictors shows that a simple Markov predictor can outperform its legacy counterparts in most of the cases. It achieves a mean accuracy of 62%, but it relies heavily on the historical data and can hardly have an enhancement from knowing individual whereabouts. Our proposed solutions can achieve a typical accuracy of 70%, which outperforms all the legacy ones and have a 1% − 5% degree of improvement by learning individual whereabouts.
Document type :
Complete list of metadata

Cited literature [50 references]  Display  Hide  Download
Contributor : Guangshuo CHEN Connect in order to contact the contributor
Submitted on : Friday, February 16, 2018 - 11:45:45 AM
Last modification on : Saturday, June 25, 2022 - 7:41:05 PM


Files produced by the author(s)


  • HAL Id : hal-01675573, version 2


Guangshuo Chen. Spatiotemporal Individual Mobile Data Traffic Prediction. [Technical Report] RT-0497, INRIA Saclay - Ile-de-France. 2018. ⟨hal-01675573v2⟩



Record views


Files downloads