marcelcaraciolo

class FriendsRecommender(MRJob):

    def steps(self):
        return [self.mr(self.map_input, self.count_number_of_friends),
                self.mr(self.count_max_of_mutual_friends,
                    self.top_recommendations)]

    def map_input(self, key, line):
        '''
        Compute a cartesian product using nested loops

marcelcaraciolo

from vectorSimilarities import VectorSimilarities


class BookSimilarities(VectorSimilarities):

    def input(self, key, line):
        user_id, item_id, rating = line.split(';')
        yield item_id, (user_id, float(rating))

marcelcaraciolo

from vectorSimilarities import VectorSimilarities


class BookSimilarities(VectorSimilarities):

    def input(self, key, line):
        user_id, item_id, rating = line.split(';')
        yield item_id, (user_id, float(rating))

marcelcaraciolo

from vectorSimilarities import VectorSimilarities


class MovieSimilarities(VectorSimilarities):

    def input(self, key, line):
        user_id, item_id, rating = line.split('|')
        yield item_id, (user_id, float(rating))

marcelcaraciolo

class VectorSimilarities(MRJob):

    def steps(self):
        return [self.mr(self.input,
                         self.group_by_user_rating),
                self.mr(None, self.count_ratings_users_freq),
                self.mr(self.pairwise_items, self.calculate_similarity),
                self.mr(self.calculate_ranking, self.top_similar_items)
                ]

marcelcaraciolo

272263;Harry Potter and the Goblet of Fire (Book 4);8
272786;Harry Potter and the Chamber of Secrets (Book 2);8
272786;Harry Potter and the Prisoner of Azkaban (Book 3);9
272786;Harry Potter and the Goblet of Fire (Book 4);9
272786;Harry Potter and the Sorcerer's Stone (Harry Potter (Paperback));9
272810;Harry Potter and the Chamber of Secrets (Book 2);8
272810;Harry Potter and the Prisoner of Azkaban (Book 3);9
272810;Harry Potter and the Goblet of Fire (Book 4);10
272810;Harry Potter and the Order of the Phoenix (Book 5);10
272810;Harry Potter and the Sorcerer's Stone (Book 1);8

marcelcaraciolo

#-*-coding: utf-8 -*-

'''
 Given a dataset of movies and their ratings by different
 users, how can we compute the similarity between pairs of
 movies?

 This module computes similarities between movies
 by representing each movie as a vector of ratings and
 computing similarity scores over these vectors.

marcelcaraciolo

#-*-coding: utf-8 -*-

'''
This module computes the number of movies rated by each
user.

'''

__author__ = 'Marcel Caraciolo <[email protected]>'

marcelcaraciolo

def jaccard(users_in_common, total_users1, total_users2):
    '''
    The Jaccard Similarity between 2 two vectors
        |Intersection(A, B)| / |Union(A, B)|
    '''
    union = total_users1 + total_users2 - users_in_common

    return (users_in_common / (float(union))) if union else 0.0

marcelcaraciolo

def regularized_correlation(size, dot_product, rating_sum, \
            rating2sum, rating_norm_squared, rating2_norm_squared,
            virtual_cont, prior_correlation):
    '''
    The Regularized Correlation between two vectors A, B

    RegularizedCorrelation = w * ActualCorrelation + (1 - w) * PriorCorrelation
        where w = # actualPairs / (# actualPairs + # virtualPairs).
    '''
    unregularizedCorrelation = correlation(size, dot_product, rating_sum, \