Expand Friend Circle Using Yelp Data

• The prediction task here is to find/recommend new friend(s) using power of data & social graph
  • Based on the users profile and their current friends, their reviews, their connections, etc
• Thereby increasing the user base of a business (and hence the reach)
• The friend prediction helps to improve the business top-line
• The data that lends itself for modelling, correlation illustrate the same.

• UserA and User B: The two user pair of friends and not-friends.
• shortestPathAB: The shortest path between the two users, this is graph based
• TotalFriendsCountA & B: A measure of extrovert-ism of users.
  
• Review Count A & B: A measure of extrovert-ism of users.
• isFriend: The outcome variable for the classification algorithms

Model Data Preparation

• Random samples of User who are not friends based on the available data in user for the city (isFriend = “N”)
• The data is split into training and test set
• The training data accuracy and the test set accruacy are used to measure the output.

Methods Used

• Random Forest: It is tree based ensemble method with 3 iterations and with cross validations.
• Gradient Boosted Tree Model: It is boosted tree based ensemble method with same 3 iterations and with cross validations.

• Test Set Accuracy: The test set accuracy is better with Gradient boosted tree with 76.6%.
• Observation 1: We are able to identify a user pair could be a friend or not based on the available & dervied.
  • Tips, and check-ins will help in identifing potential pairs
• Observation 2: Help Business & Yelp to have a cohesive communities.