Transcript results - UCLA Computer Science

An Investigation into
Guest Movement in
the Smart Party
Jason Stoops ([email protected])
Faculty advisor: Dr. Peter Reiher
Outline
Project Introduction
 Key metrics and values
 Mobility Models, Methods of Testing
 Results
 Analysis

What is the Smart Party?
Ubiquitous computing application
 Someone hosts a gathering
 Guests bring wireless-enabled devices
 Devices in the same room cooperate to
select and supply media to be played
 Songs played in a room represent tastes
of guests present in that room

Project Motivation
Are there ways to move between rooms in
the party that can lead to greater
satisfaction in terms of music heard?
 Can we ultimately recommend a room for
the user?
 What other interesting tidbits about the
Smart Party can we come up with along
the way?

Smart Party Simulation Program



Basis for evaluating mobility models (rules of
movement).
Real preference data from Last.FM is used.
Random subsets of users and songs chosen
 Many
parties with same conditions are run with
different subsets to gather statistics about the party.

Initial challenge: extend existing simulation to
support multiple rooms.
Metrics

Satisfaction: based on 0-5 “star” rating
 Rating
determined by play count
 Exponential scale: k-star rating = 2k satisfaction
 0-star rating = 0 satisfaction (song unknown)

Fairness: distribution of satisfaction
Coefficient – usually used for measuring
distribution of wealth in a population.
 In Smart Party, wealth = satisfaction.
 Ratio between 0 to 1, lower is more fair.
 Gini
Key values

History Length
 Number
of previously heard songs the user device
will track.
 Used to evaluate satisfaction with current room

Satisfaction Threshold
 Used
as a guide for when guest should consider
moving.
 If average satisfaction over last history-length songs
falls below sat-threshold, guest considers moving.
Mobility Models Tested
No movement
 Random movement
 Threshold-based random movement
 Threshold-based to least crowded room
 Threshold-based, population weighted
 Threshold-based, highest satisfaction

Test Procedure
Round 1: Broad testing to find good values
for history length and satisfaction
threshold for each model. (25 iterations)
 Round 2: In-depth evaluation of model
performance using values found above.
(150 iterations)
 Ratio of six guests per room maintained

Round 1 Results
Model
History Length
Threshold
No Movement
n/a
n/a
Random
n/a
n/a
Threshold Random
4
1
Threshold Least
Crowded
4
1
Threshold Random, 5
Population Weighted
0.5
Threshold Highest
Satisfaction
2.25
2
Round 2: Satisfaction Overview
Median Overall Satisfaction
25th, 75th quartiles shown
250
Satisfaction
200
150
NOMOVE
THRESHOLD LEAST
CROWDED
THRESHOLD RANDOM POP
WEIGHTED
100
RANDOM
THRESHOLD RANDOM
THRESHOLD HIGHEST SAT
50
0
30 guests / 5 rooms
18 guests / 3 rooms
90 guests / 15 rooms
60 guests / 10 rooms
Round 2: Fairness Overview
Median Overall Fairness
25th, 75th quartiles shown, lower is better
0.45
0.4
0.35
0.2
NOMOVE
THRESHOLD LEAST
CROWDED
THRESHOLD RANDOM POP
WEIGHTED
RANDOM
0.15
THRESHOLD RANDOM
THRESHOLD HIGHEST SAT
0.3
Fairness
0.25
0.1
0.05
0
30 guests / 5 rooms
18 guests / 3 rooms
90 guests / 15 rooms
60 guests / 10 rooms
Topics for Analysis
Moving is better than not moving
 Party stabilization?
 Initial room seeking
 Population-based models perform poorly
 Satisfaction-based model performs well

Moving Versus Not Moving

Movement “stirs”
party, making
previously
unavailable songs
accessible
Songs users have in
common changes
with movement,
depleted slower.
Random vs. No Move, Median Overall Satisfaction
25th, 75th quartiles shown
200
180
160
140
Satisfaction

18 Guests / 3
Rooms
30 Guests / 5
Rooms
60 Guests / 10
Rooms
120
100
80
60
40
20
0
NOMOVE
RANDOM
Party stabilization?


Do users find “ideal
rooms” and stop
moving?
No! Some movement
is always occurring.
Cause: Preferences
are not static, they
evolve over time.
Room Changes per Guest over Time
Threshold Highest Sat, 30 guests / 5 rooms
0.9
Movement Probability

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
5
10
15
Round
20
25
30
35
Initial room seeking


90% of guests move
after round 1
Guests have some
information to go on
after one song plays.
Guests that like the
first song in a room
likely have other
songs in common.
Room Changes per Guest over Time
Threshold Highest Sat, 60 guests / 10 rooms
1
0.9
0.8
Movement Probability

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
5
10
15
Round
20
25
30
35
Initial room seeking, cont.
Round-by-round Satisfaction
60 Guests in 10 Rooms
8
7
6
Satisf action
5
NOMOVE
RANDOM
THRESHOLD RANDOM
4
THRESHOLD HIGHEST SAT
3
2
1
0
0
5
10
15
20
25
30
35
Round


In satisfaction-based model, peak is in round 2
All other models peak in round 1.
Population-based models


Worse than choosing
a room at random!
Weighted model
performed better as
weighting approached
being truly random.
However, still better
than not moving at all.
Median Overall Satisfaction
25th, 75th quartiles shown
200
180
160
140
Satisfaction

120
100
80
60
40
20
0
18 guests / 3 rooms
NOMOVE
THRESHOLD
LEAST
CROWDED
30 guests / 5 rooms
THRESHOLD
RANDOM POP
WEIGHTED
RANDOM
Satisfaction based model


Informed movement
better than random
movement.
Greater advantage as
more rooms are
added.
Short history length
(two songs) used
since history goes
“stale”.
Median Overall Satisfaction
25th, 75th quartiles shown
250
200
Satisfaction

150
100
50
0
30 guests / 5 rooms
90 guests / 15 rooms
18 guests / 3 rooms
60 guests / 10 rooms
NOMOVE
RANDOM
THRESHOLD
RANDOM
THRESHOLD
HIGHEST SAT
Conclusion
Room recommendations are a feasible
addition to the Smart Party User Device
Application.
 Recommendations based on songs played
are more valuable than those based on
room populations.
 Movement is a key part of the Smart Party.

Acknowledgements

At the UCLA Laboratory for Advanced Systems
Research:
 Dr. Peter Reiher
 Kevin Eustice
 Venkatraman Ramakrishna
 Nam

Nguyen
For putting together the UCLA CS
Undergraduate Research Program
 Dr. Amit Sahai
 Vipul Goyal
References




Eustice, Kevin; Ramakrishna, V.; Nguyen, Nam; Reiher, Peter, "The Smart
Party: A Personalized Location-Aware Multimedia Experience," Consumer
Communications and Networking Conference, 2008. CCNC 2008. 5th IEEE
, vol., no., pp.873-877, 10-12 Jan. 2008
Kevin Eustice, Leonard Kleinrock, Shane Markstrum, Gerald Popek,
Venkatraman Ramakrishna, Peter Reiher . Enabling Secure Ubiquitous
Interactions, In the proceedings of the 1st International Workshop on
Middleware for Pervasive and Ad-Hoc Computing (Co-located with
Middleware 2003), 17 June 2003 in Rio de Janeiro, Brazil.
Gini, Corrado (1912). "Variabilità e mutabilità" Reprinted in Memorie di
metodologica statistica (Ed. Pizetti E, Salvemini, T). Rome: Libreria Eredi
Virgilio Veschi (1955).
Audioscrobbler. Web Services described at
http://www.audioscrobbler.net/data/webservices/