This document discusses opportunistic analytics for modeling human activity. It presents a methodology that involves collecting and combining data from multiple sources to increase information density, segmenting and profiling behaviors, and inferring activity trajectories. Two case studies are described: one using location-aware social media data to identify 10 activity types, and another using in-home internet activity to map applications to 8 activities. The studies demonstrate enhancing data density and predicting future activity patterns with over 70% accuracy.
3. Research by Development, KISS Principle, Top Down and End-to-End approach.
Research Objective
Research Methodology
Design and Development of Large Scale Distributed Systems for Next Generation Communication and Data Driven
Telecom Services.
Scalable Systems
Department
31. Dataset
LBSN - Foursquare and Twitter - Traces
Custom collected Geo-Fenced Tweets, and Foursquare Tweets
are analyzed to construct activity trajectory.
825 Users
79431 Check-ins
1 Year
157806 Geo-Tweets
30 KM
Geo tagged Tweets with
embedded Foursquare
check-in URL
Longitude, Latitude,
Timestamp, Location
Name, and Category
Dataset
32. Location + Time + Venue Type = Activity
“Primary FourSqaure venue categories
are analyzed with 325 locations to
extract 10 distinct activity types”.
39. Dataset : Project LeYLab
In-Home Internet Activity Traces
Living Lab for Fiber based Services in the City of Kortrijk, Belgium.
ALU 7750 Service Router with Report and Analysis Manager (RAM) was used in the backbone.
86 Households
75 Applications
60 Days
9288000 Data Points
41. Accumulated activity footprint of a representative household, activity is spread over through
out the day, with higher engagements during evenings
Activity Trajectory
0
5
10
15
20
25
6 AM 8 AM 10 AM 12 PM 2 PM 4 PM 6 PM 8 PM 10 PM 12 AM 2 AM 4 AM
Web Communication
Soical Networking
Online Gaming
Home Working
Online Shopping
Video Watching
Time of the Day
NoofDays
42. We have observed an inverse relationship between application usage frequency and
corresponding traffic load.
Accordingly, we model activity using interaction frequency and temporal regularity. This
measure identifies how a household engages with a distinct activity.
Recurrence Measure
44. Trajectory Prediction Algorithm
The algorithm predicts activity patterns of future hour slots of current day by matching
patterns of similar days in the past.
45. Prediction Performance
60% of households activities can be predicted accurately 70% of times.
CumulativeDistributionFunction(CDF)
0
0.2
0.4
0.6
0.8
1.0
0 0.2 0.4 0.6 0.8 1.0
F-Measure
46. Opportunistic Observation
By observing an individual’s engagement with semantically rich applications annotated with
temporal and spatial information, we can infer an individuals activity.
Key Points
Density Enhancement
Multiple Traces can be combined by co-relating their spatio-temporal properties to increase
information density.
Trajectory Prediction
The algorithm predicts activity patterns of future hour slots of current day by matching
patterns of similar days in the past.