Upload
michael-steurer
View
186
Download
0
Tags:
Embed Size (px)
DESCRIPTION
Recent research has shown that digital online geo- location traces are new and valuable sources to predict social interactions between users, e.g. , check-ins via FourSquare or geo-location information in Flickr images. Interestingly, if we look at related work in this area, research studying the extent to which social interactions can be predicted between users by taking more than one location-based knowledge source into account does not exist. To contribute to this field of research, we have collected social interaction data of users in an online social network called My Second Life and three related location-based knowledge sources of these users (monitored locations, shared locations and favored locations), to show the extent to which social interactions between users can be predicted. Using supervised and unsupervised machine learning techniques, we find that on the one hand the same location-based features (e.g. the common regions and common observations) perform well across the three different sources. On the other hand, we find that the shared location information is better suited to predict social interactions between users than monitored or favored location information of the user.
Citation preview
Predicting Social InteractionsBased on Different Sources of Location-based Knowledge
Michael Steurer - [email protected], Graz University of TechnologyChristoph Trattner - [email protected], Graz University of TechnologyDenis Helic - [email protected], Graz University of Technology
What Did We Do?
Predciting Social Interactions Using Location-Based Sources
Predict Interactions in Social Network
Three Different Sources of Location Data
Postings, Comments, Loves
Similar to Facebook, G+
Monitored Locations
Shared Locations
Favoured Locations
2/23
Second Life
Predciting Social Interactions Using Location-Based Sources
source: http://notizen.typepad.com/aus_der_provinz/sl061018_001_1.jpg
3/23
Online Social NetworkText-Interaction Data
Online Social Network
Predciting Social Interactions Using Location-Based Sources 5/23
Collected Data
Predciting Social Interactions Using Location-Based Sources
Online Social Data
Groups and Interests
152,509 Unique Users
1,084,002 Postings (Text Messages, Snapshots)
459,734 Comments
1,631,568 Loves
285,528 Unique Groups
15.51 Groups per User on Average
6.5 Interests per User on Average
6/23
Location-based InformationUsers Position Data
1) Monitored Locations
Predciting Social Interactions Using Location-Based Sources 8/23
Collected Data
Predciting Social Interactions Using Location-Based Sources
Event Data
Location-based Social Data
12 Months Starting in March 2012
Working Hours 24/7
4,105 Unique Locations
19 Million Data Samples
410,619 Unique Users
9/23
2) Shared Locations
Predciting Social Interactions Using Location-Based Sources 10/23
Collected Data
Predciting Social Interactions Using Location-Based Sources
Compare to Check-ins, e.g. FourSquare
Shared from "In-World"
Harvested Locations
45,835 User Profiles
496,912 Snapshots
13,583 Unique Locations
11/23
3) Favoured Locations
Predciting Social Interactions Using Location-Based Sources 12/23
Collected Data
Predciting Social Interactions Using Location-Based Sources
Extracted from Profiles
Limitations
Favoured Locations
10 per User
Enhanced with Picture and Text
191,610 User Profiles
811,386 Picks
25,311 Unique Locations
13/23
Predict Text-InteractionsUse Location Information
Network Setup
Predciting Social Interactions Using Location-Based Sources
Create Networks from Data
Online Social Network
Enhance with Location-based Data
15/23
Feature Modeling
Predciting Social Interactions Using Location-Based Sources
Compute User Relation
Common and Total Locations (Jaccard's Coefficient)
Entropy Common Locations
User Count Common Locations
Frequency Common Locations
16/23
Experiment Setup
Predciting Social Interactions Using Location-Based Sources
Feature Modeling
Prediction Task
Unsupervised Learning with Ranked Lists
Information Gain for Single Features
Binary Classification Problem
Supervised Learning Algorithms
Logistic Regression, Random Forest, SVM
17/23
Results
Analysis of Homophily
Predciting Social Interactions Using Location-Based Sources
(*p < 0.1, **p < 0.01, and ***p < 0.001)
19/23
Predict Interactions
Predciting Social Interactions Using Location-Based Sources 20/23
Conclusions
Conclusions
Predciting Social Interactions Using Location-Based Sources
User-Pairs with Interactions
Results of the Prediction
More Common and Total Locations
Locations have Less Entropy, Frequency, and User-Count
Common Locations, Jaccard
Shared > Monitored > Favoured
Same Characteristics Among Algorithms
Logistic Regression was Best
22/23
Predicting Social InteractionsBased on Different Sources of Location-based Knowledge
Michael Steurer - [email protected], Graz University of TechnologyChristoph Trattner - [email protected], Graz University of TechnologyDenis Helic - [email protected], Graz University of Technology