Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

Data Analysis in LTE

Networks
Mobile Radio Networks 2021
Prof. Antonio Capone
Eng. Andrea Pimpinella
Project Proposal N1.
• Develop a Supervised Machine Learning Classifier f ! able to
predict user’s satisfaction relatively to the usage of YouTube
Service. The Classifier will take as input Crowdsourcing Data (i.e.,
data measured directly at users terminals) and will output the
corresponding users satisfaction class.

𝑿
ML
𝑺"
Algorithm

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 2


Project Proposal N1.
Tasks

• Project Tasks:
1. Provide a qualitative/descriptive analysis of the data
2. Augment the basic model, justifying the reasons for which the added
features or the chosen mathematical transformation are useful for prediction
3. Express the prediction performance with AUC metric

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 3


Project Tools
Python Template

Connect to:

PYTHON TEMPLATE FOR PROJECT PROPOSAL N.1

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 4


Drivers for Mobile Network Operators
The goals of MNOs

• Ensure high Quality of Experience to customers: Model the


• Deliver good Quality of Service (QoS) mapping between
• Attract as many customers as possible QoS and QoE
• Minimise the Churn Rate

• Improve efficiency of Network Management Leverage Big Data to


processes: improve the accuracy
• Optimise Network Dimensioning and Resources of mobile network
Allocation strategies dynamics modeling
• Discover issues in the network (Network
Monitoring)
Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 5
Model Mobile Traffic Dynamics

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 6


Mobile Network Monitoring

• Mobile Network Operators (MNOs) work to recognize under-performing network


layers or elements, with the aim of maximizing the efficiency of their networks

• To fulfill this goal, cellular operators usually undertake detailed analysis on


heterogeneous types of on-field collected network data

• In particular, the collection of on-field measurements over large time periods


(months or years) is crucial to recognize trends happening in the network.

• This kind of analysis are tipically done considering data measured at Base Station
premises through ad-hoc installed Network Counters

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 7


Manage Network Data
Dimensionality of the scenario

• A city-wide mobile network can count up to several thousands of different


network sites

• Considering a data sampling time of 1 hour, one month of data for a single KPI
and a single network site counts 720 data samples

• E.g., a dataset which collects hourly traffic loads served by the eNodeBs of a LTE
network of medium size (1500 network sites) in 1 month of activity counts more
than 1M data points

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 8


Manage Network Data
DL Traffic Trace

• 1 Month DL Traffic
Data (January 2020)

• Single eNodeB
Kbit

• Time series of
24x30=720 hourly
samples

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 9


Clustering
How to reduce the complexity

• It has been shown that there is a strong relationship between mobile users
network activity and the characteristics of the urban area where the
activity takes place

• In other words, it is possible to group network sites according to the spatial


and temporal dynamics of the served traffic

• This is called Clustering

• Rationale: the network activity of the eNodeBs of the same cluster can be
explained with the same model

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 10


How To Cluster eNodeBs?
Build a Traffic Signature

• We need to identify an object which represents the traffic activity


of a eNodeB in the observation period, i.e., the eNodeB traffic
signature

• Signature support: 𝜹 = { 𝑊𝑒𝑒𝑘 𝐷𝑎𝑦, 𝑆𝐴𝑡𝑢𝑟𝑑𝑎𝑦, 𝑆𝑈𝑛𝑑𝑎𝑦}

• 𝒅! = « 𝑆𝑒𝑡 𝑜𝑓 𝑑𝑎𝑦𝑠 𝑑 𝑜𝑓 𝑡𝑦𝑝𝑒 𝛿 ∈ 𝜹» à 𝒅"# is the set of all


Mondays, Tuesdays,…, Fridays contained in the observation
period

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 11


How To Cluster eNodeBs?
Build a Traffic Signature

Kbit

Week Days SAturdays SUndays

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 12


How To Cluster eNodeBs?
Median Weekly Signature

2 • Median Profile of DL Traffic at the


considered eNodeB (here, in 1
1.5 Month)

1
• It is composed of 5x WD + 1X SA +
0.5 1X SU Median Profiles

0
• Clustering is performed trhough a K-
-0.5 Means Algorithm over the set of all
Median Weekly Signatures
-1
Mon. Tue. Wed. Thu. Fri. Sat. Sun.

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 13


A Typical Clustering Configuration
RESIDENTIAL CLUSTER
2
• The output of the
Norm. Traffic

clustering algorithm is the


«centroids» of the clusters,
0

-2
i.e., the MWSs which
Mon. Tue. Wed. Thu.
BUSINESS CLUSTER
Fri. Sat. Sun. represent all the eNodeBs
of the same cluster
2
Norm. Traffic

-2
Mon. Tue. TRANSPORTATION
Wed. Thu. CLUSTER
Fri. Sat. Sun.
2
Norm. Traffic

-2
Mon. Tue. Wed. Thu. Fri. Sat. Sun.

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 14


Project Proposal N.2
Description

• Study the impact of the lock-down due to COVID-19 pandemic on


the 800 MHz frequency layer of Vodafone LTE network in the
area of Milan

• To do so, you can exploit Radio Access Network data relatively


collected between January 1st and March 31st 2020 (full lock-
down in Italy started on 9th March 2020)

• In particular, consider the period before 16° February 2020 as


«Covid-Free» observation period
Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 15
Project Proposal N.2
Objectives

• Analyse data to recognise weekly trends, providing KPIs statistics


(e.g., daily or weekly median, average, etc.) to:

1. Recognise (and differentiate) residential and non-residential


network areas (e.g., through clustering)
2. Study the variation of residential and non-residential network
cells performance due to lock down (e.g., congestion in
residential areas has increased, mobility in the network has
reduced, etc.)

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 16


The Dataset
Project Proposal N.2

• 16 Radio Access KPIs of 3 Categories (Accessibility, Mobility and Usage)


• Data samples are provided with Hourly granularity
• Network Cells Identifiers and Exact Location

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 17


Project Tools
Python Template

Connect to:

PYTHON TEMPLATE FOR PROJECT PROPOSAL N.2

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 18


References
• [1] A. Okic, A. E. C. Redondi, I. Galimberti, F. Foglia and L. Venturini, "Analyzing
Different Mobile Applications in Time and Space: a City-Wide Scenario," 2019 IEEE
Wireless Communications and Networking Conference (WCNC), Marrakesh, Morocco,
2019, pp. 1-6.
• [2] Furno, Angelo, et al. "A tale of ten cities: Characterizing signatures of mobile traffic
in urban areas." IEEE Transactions on Mobile Computing 16.10 (2016): 2682-2696.
• [3] A. Pimpinella, F. Di Giusto, A.E.C. Redondi, "Cluster-Based Forecasting of Busy-Hour
Downlink Traffic in Cellular Networks"
• [4] Google LLC "Google COVID-19 Community Mobility Reports."
https://www.google.com/ covid19/mobility/

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 19


Rules for Projects Delivery
1. Choose one project between the two proposals
2. Maximum N. of Points: 5
3. Delivery: ppt. Or .pdf (essay like) which describes your approach
and results
4. Deadline: Before Last September Exam Call (NB. For all those
who need to take course credits for scolarship confirmation, be
aware of your own deadlines!)
5. Projects can be tackled either individually or by group (Max. 2
People). If by groupp, clarify the contribute of each member.
6. Contacts: andrea.pimpinella@polimi.it

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella 20


Questions?

Thank you for your kind attention!

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella


BACKUP

Data Analysis in LTE Networks – MRN2021 - Andrea Pimpinella

You might also like