Step
4.
Our first Recommender
• The input is an ARFF file generated from
the Anduin bookstore database (‘bookStore.arff’)
• Each data-instance is a user, and the
attributes represent the different books, the Boolean value tells us whether or
not this user has bought that book.
• We need the active user description (In
file ActiveUser.arff)
• We need to find 5 nearest neighbor
(critics) using LinearNNSearch class
• We need distance of the active user from
each neighbor, which we will convert to a similarity score: sim=1/distance
• We are going to output 3 top
recommendations
Try it out:
• The code is in file ‘WekaRecommender.java’
• The training set is in bookStore.arff
• The active user is in ActiveUser.arff