Multi-modal Mining and Modeling of Big Mobile Networks Based on Users Behavior and Interest
📝 Abstract
Usage of mobile wireless Internet has grown very fast in recent years. This radical change in availability of Internet has led to communication of big amount of data over mobile networks and consequently new challenges and opportunities for modeling of mobile Internet characteristics. While the traditional approach toward network modeling suggests finding a generic traffic model for the whole network, in this paper, we show that this approach does not capture all the dynamics of big mobile networks and does not provide enough accuracy. Our case study based on a big dataset including billions of netflow records collected from a campus-wide wireless mobile network shows that user interests acquired based on accessed domains and visited locations as well as user behavioral groups have a significant impact on traffic characteristics of big mobile networks. For this purpose, we utilize a novel graph-based approach based on KS-test as well as a novel co-clustering technique. Our study shows that interest-based modeling of big mobile networks can significantly improve the accuracy and reduce the KS distance by factor of 5 comparing to the generic approach.
💡 Analysis
Usage of mobile wireless Internet has grown very fast in recent years. This radical change in availability of Internet has led to communication of big amount of data over mobile networks and consequently new challenges and opportunities for modeling of mobile Internet characteristics. While the traditional approach toward network modeling suggests finding a generic traffic model for the whole network, in this paper, we show that this approach does not capture all the dynamics of big mobile networks and does not provide enough accuracy. Our case study based on a big dataset including billions of netflow records collected from a campus-wide wireless mobile network shows that user interests acquired based on accessed domains and visited locations as well as user behavioral groups have a significant impact on traffic characteristics of big mobile networks. For this purpose, we utilize a novel graph-based approach based on KS-test as well as a novel co-clustering technique. Our study shows that interest-based modeling of big mobile networks can significantly improve the accuracy and reduce the KS distance by factor of 5 comparing to the generic approach.
📄 Content
Multi-modal Mining and Modeling of Big Mobile Networks Based on Users Behavior and Interest Saeed Moghaddam, Ahmed Helmy Computer and Information Science and Engineering Department University of Florida moghaddam@ufl.edu, helmy@ufl.edu
Abstract— Usage of mobile wireless Internet has grown
very fast in recent years. This radical change in availability of
Internet has led to communication of big amount of data over
mobile networks and consequently new challenges and
opportunities for modeling of mobile Internet characteristics.
While the traditional approach toward network modeling
suggests finding a generic traffic model for the whole network,
in this paper, we show that this approach does not capture all
the dynamics of big mobile networks and does not provide
enough accuracy. Our case study based on a big dataset
including billions of netflow records collected from a campus-
wide wireless mobile network shows that user interests acquired
based on accessed domains and visited locations as well as user
behavioral groups have a significant impact on traffic
characteristics of big mobile networks. For this purpose, we
utilize a novel graph-based approach based on KS-test as well as
a novel co-clustering technique. Our study shows that interest-
based modeling of big mobile networks can significantly
improve the accuracy and reduce the KS distance by factor of 5
comparing to the generic approach.
Keywords- user interest; mobile data; traffic; big data;
co-clustering
I.
INTRODUCTION
Mobile Internet traffic has experienced a significant
growth in the recent years. Different types of Internet-enabled
mobile devices are getting more and more popularity and
wireless Internet access infrastructures are growing faster
than ever. The emergence of this radical change in
availability of Internet raises a new need for modeling of
Internet characteristics in big mobile networks. A traffic
model in general is a model that can be used to regenerate the
behavior of a real traffic stream. A major application of
traffic models is in predicting the behavior of traffic as it
passes through a network. The common approach toward
traffic modeling is to find a generic model for the whole
network.
Although,
such
models
provide
good
approximations for the old wired Internet, but several studies
have shown that they do not fit the dynamics of wireless
networks. For example, [1] characterizes the wireless traffic
in different locations and shows that the dynamics of network
follow a similar model but with different parameters.
However, such models are generally based on small datasets
of WLAN activities (e.g. 25000 flows a day), which are far
from the full scale of dynamics in current big mobile
networks (e.g., our dataset includes over 100 million flows
per day). Moreover, most previous works have not studied the
characteristics of big mobile networks based on user behavior
and interests which can be acquired based on accessed
domains (e.g. ‘cnn’) or visited locations (e.g. ‘cinema’).
Interest-based and behavior-aware modeling of big mobile
network traffic can be beneficial to the realistic design of
applications, protocols and services (e.g. for resource
allocating or content caching).
In this paper, we present a novel modeling approach
based on our earlier work on graph-based traffic analysis of
domains and locations [2] and also introduce a novel
technique for analysis and modeling of multi-modal user
behavioral groups using a co-clustering approach. Our
campus-wide case study shows that domains, locations and
users have specific traffic characteristic that can also form
groups with distinct characteristics. In our study, we
investigate interest-based characteristics by partitioning the
Internet traffic based on domains, buildings and user
behavioral groups and analyzing the traffic characteristics
using KS (Kolmogorov-Smirnov) test [3].
This work has the following key contributions:
1.
We provide a novel interest-based traffic modeling
technique for big mobile networks based on accessed
domains (top 100 active domains) and visited locations (68
different buildings) across a campus with more than 32000
users. The studied dataset is one of the largest wireless
mobile network traffic traces (including around 100 million
records per day).
2. We provide a systematic method to discover
similarities and differences between the traffic distributions
of different domains or locations. We show how a novel
graph-based technique can be applied to identify groups of
domains or locations with distinct traffic characteristics.
3. We provide a novel technique to discover multi-modal
user behavioral groups based on both domains and locations
visitations. We show how a co-clustering technique based on
information theory can be utilized to identify behavioral
groups with distinct traffic characteristics.
4. We show that the propos
This content is AI-processed based on ArXiv data.