YOU ARE WHAT YOU EAT (AND DRINK): IDENTIFYING CULTURAL BOUNDARIES BY ANALYZING FOOD AND DRINK HABITS IN FOURSQUARE Presenter: LEUNG Pak Him
METHODS USED TO ANALYZE CROSS- CULTURAL DIFFERENCES Traditional method Surveys New method in this paper Foursquare check-ins
PROCEDURES 1) Map food and drink related check-ins2) Identify particular individual preferences 3) Show how to analyze this information assess the cultural distance 4) Apply a simple k-means clustering technique to draw boundaries
CULTURAL BOUNDARIES Homophily Social Influence Cultural Boundaries
TRADICTIONAL METHOD CONSTRAINTS
BIGGEST CHALLENGE IN THE ANALYSIS Problem: No appropriate empirical data to use Problem: No appropriate empirical data to use Solution: data collected from questionnaires filled during face-to-face interviews Solution: data collected from questionnaires filled during face-to-face interviews
CONSTRAINTS IN USING SURVEY DATA 1) costly and do not scale up 2) provide only static information
NEW METHOD
REQUIREMENTS FOR USING NEW METHOD 1) Associate a user to its location 2) Extract a finite set of preferences 3) Map users’ actions into the preferences
MAPPING PREFERENCES
DATA DESCRIPTION Eight main venue categories Eight main venue categories Sub-categories Sub-categories Spans a single week of April 2012 Spans a single week of April 2012 Grouped relevant subcategories into three classes
FREQUENCY OF CHECK-INS OF THE THREE ANALYZED CLASSES ClassDrinkFast FoodSlow Food Check-ins279,650410,592394,042 Unique venue106,152193,541198,565 Unique users162,891230,846231,651 No. of subcategories212753
MAPPING FOURSQUARE DATA INTO USER PREFERENCES m =101 features m =101 features F = a vector of 101 attributes with binary representation F = a vector of 101 attributes with binary representation Finite set of preferences Finite set of preferences Map users’ action Map users’ action Associate a user with a location Associate a user with a location
CULTURAL SIMILARITIES
EXAMPLE NETWORKS IN THE PAPER
ANALYSIS OF THE EXAMPLE NETWORKS % of people satisfying “s” +1 : people living in the same region tend to be similar -1 : people living in the same region tend to be different
SPATIAL CORRELATIONS Goal : Define a set of features that are able to characterize the cultural preferences of a given geographical area Goal : Define a set of features that are able to characterize the cultural preferences of a given geographical area 3) Calculate Pearson’s correlation for different area vectors
CORRELATION MATRICES BETWEEN COUNTRIES
CORRELATION MATRICES BETWEEN CITIES
WITHIN BORDER ANALYSIS
CORRELATION MATRICES
TEMPORAL ANALYSIS 1) Count the number of check-ins per hour2) Group days into weekdays and weekends3) Normalize the combined number
RESULT - 1
IDENTIFYING CULTURAL BOUNDARIES
CLUSTERING REGIONS 2) Apply the Principal Component Analysis3) Apply k-means algorithm
RESULT
Q & A