Skip to content
This repository

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
branch: master
Fetching contributors…

Octocat-spinner-32-eaf2f5

Cannot retrieve contributors at this time

file 326 lines (258 sloc) 10.787 kb
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
#-*-coding: utf-8 -*-

'''
Given a dataset of ratings, how can we compute the similarity
between pairs of items ?

Subclasses from this module that provide a concrete implementation
of the input (in the form of a tuple stream containing: a user, the
item being rated, and the numeric rating of the item by the user)
will calculate the similarities of items.

Each item is represented as a (sparse) vector of all its ratings.
Similarity measures (such as correlation, cosine and Jaccard) will
be applied to these vectors.

'''
__author__ = 'Marcel Caraciolo <caraciol@gmail.com>'

from mrjob.job import MRJob
from math import sqrt

try:
    from itertools import combinations
except ImportError:
    def combinations(iterable, r):
        """
Implementation of itertools combinations method.
Re-implemented here because of import issues
in Amazon Elastic MapReduce. Just easier to do this than bootstrap.
More info here:
http://docs.python.org/library/itertools.html#itertools.combinations

Input/Output:

combinations('ABCD', 2) --> AB AC AD BC BD CD
combinations(range(4), 3) --> 012 013 023 123
"""
        pool = tuple(iterable)
        n = len(pool)
        if r > n:
            return
        indices = range(r)
        yield tuple(pool[i] for i in indices)
        while True:
            for i in reversed(range(r)):
                if indices[i] != i + n - r:
                    break
            else:
                return
            indices[i] += 1
            for j in range(i + 1, r):
                indices[j] = indices[j - 1] + 1
            yield tuple(pool[i] for i in indices)


#Parameters to regularize correlation
PRIOR_COUNT = 10
PRIOR_CORRELATION = 0

#FILTERS to speed up computation and reduce noise
#Subclasses should probably override these, based on actual data.
MIN_NUM_RATERS = 2
MAX_NUM_RATERS = 10000
MIN_INTERSECTION = 0


class SemicolonValueProtocol(object):

    # don't need to implement read() since we aren't using it

    def write(self, key, values):
        return ';'.join(str(v) for v in values)


class VectorSimilarities(MRJob):

    OUTPUT_PROTOCOL = SemicolonValueProtocol

    def steps(self):
        return [
            self.mr(mapper=self.input, reducer=self.group_by_user_rating),
            self.mr(reducer=self.count_ratings_users_freq),
            self.mr(mapper=self.pairwise_items,
                    reducer=self.calculate_similarity),
            self.mr(mapper=self.calculate_ranking,
                    reducer=self.top_similar_items)]

    def configure_options(self):
        super(VectorSimilarities, self).configure_options()
        self.add_passthrough_option(
            '--priorcount', dest='prior_count', default=10, type='int',
            help='PRIOR_COUNT: Parameter to regularize correlation')

        self.add_passthrough_option(
            '--priorcorrelation', dest='prior_correlation', default=0,
             type='int',
             help='PRIOR_CORRELATION: Parameter to regularize correlation')

        self.add_passthrough_option(
            '--minraters', dest='min_num_raters', default=3, type='int',
            help='the minimum number of raters')

        self.add_passthrough_option(
            '--maxraters', dest='max_num_raters', default=10000, type='int',
            help='the maximum number of raters')

        self.add_passthrough_option(
            '--minintersec', dest='min_intersection', default=0, type='int',
            help='the minimum intersection')

    def input(self, key, line):
        '''
Subclasses should override this to define their own input
'''
        raise NotImplementedError('Implement this in the subclass')

    def group_by_user_rating(self, key, values):
        """
Emit the user_id and group by their ratings (item and rating)

17 70,3
35 21,1
49 19,2
49 21,1
49 70,4
87 19,1
87 21,2
98 19,2

"""
        total = 0
        final = []
        for user_id, rating in values:
            total += 1
            final.append((user_id, rating))

        if total >= self.options.min_num_raters and \
           total <= self.options.max_num_raters:
            for user_id, rating in final:
                yield user_id, (key, float(rating), total)
                #yield None, '%s;%s;%.2f;%d' % (user_id, key, rating, total)

    def count_ratings_users_freq(self, user_id, values):
        """
For each user, emit a row containing their "postings"
(item,rating pairs)
Also emit user rating sum and count for use later steps.

17 1,3,(70,3)
35 1,1,(21,1)
49 3,7,(19,2 21,1 70,4)
87 2,3,(19,1 21,2)
98 1,2,(19,2)
"""
        item_count = 0
        item_sum = 0
        final = []
        for item_id, rating, ratings_count in values:
            item_count += 1
            item_sum += rating
            final.append((item_id, rating, ratings_count))

        yield user_id, (item_count, item_sum, final)

    def pairwise_items(self, user_id, values):
        '''
The output drops the user from the key entirely, instead it emits
the pair of items as the key:

19,21 2,1
19,70 2,4
21,70 1,4
19,21 1,2

This mapper is the main performance bottleneck. One improvement
would be to create a java Combiner to aggregate the
outputs by key before writing to hdfs, another would be to use
a vector format and SequenceFiles instead of streaming text
for the matrix data.
'''
        item_count, item_sum, ratings = values
        #print item_count, item_sum, [r for r in combinations(ratings, 2)]
        #bottleneck at combinations
        for item1, item2 in combinations(ratings, 2):
            yield (item1[0], item2[0]), \
                    (item1[1], item2[1], item1[2], item2[2])

    def calculate_similarity(self, pair_key, lines):
        '''
Sum components of each corating pair across all users who rated both
item x and item y, then calculate pairwise pearson similarity and
corating counts. The similarities are normalized to the [0,1] scale
because we do a numerical sort.

19,21 0.4,2
21,19 0.4,2
19,70 0.6,1
70,19 0.6,1
21,70 0.1,1
70,21 0.1,1
'''
        sum_xx, sum_xy, sum_yy, sum_x, sum_y, n = (0.0, 0.0, 0.0, 0.0, 0.0, 0)
        n_x, n_y = 0, 0
        item_pair, co_ratings = pair_key, lines
        item_xname, item_yname = item_pair
        for item_x, item_y, nx_count, ny_count in lines:
            sum_xx += item_x * item_x
            sum_yy += item_y * item_y
            sum_xy += item_x * item_y
            sum_y += item_y
            sum_x += item_x
            n += 1
            n_x = int(ny_count)
            n_y = int(nx_count)

        corr_sim = correlation(n, sum_xy, sum_x, \
                 sum_y, sum_xx, sum_yy)

        reg_corr_sim = regularized_correlation(n, sum_xy, sum_x, \
                sum_y, sum_xx, sum_yy, PRIOR_COUNT, PRIOR_CORRELATION)

        cos_sim = cosine(sum_xy, sqrt(sum_xx), sqrt(sum_yy))

        jaccard_sim = jaccard(n, n_x, n_y)

        yield (item_xname, item_yname), (corr_sim, \
                cos_sim, reg_corr_sim, jaccard_sim, n)

    def calculate_ranking(self, item_keys, values):
        '''
Emit items with similarity in key for ranking:

19,0.4 70,1
19,0.6 21,2
21,0.6 19,2
21,0.9 70,1
70,0.4 19,1
70,0.9 21,1

'''
        corr_sim, cos_sim, reg_corr_sim, jaccard_sim, n = values
        item_x, item_y = item_keys
        if int(n) > self.options.min_intersection:
            yield (item_x, corr_sim, cos_sim, reg_corr_sim, jaccard_sim), \
                     (item_y, n)

    def top_similar_items(self, key_sim, similar_ns):
        '''
For each item emit K closest items in comma separated file:

De La Soul;A Tribe Called Quest;0.6;1
De La Soul;2Pac;0.4;2

'''
        item_x, corr_sim, cos_sim, reg_corr_sim, jaccard_sim = key_sim
        for item_y, n in similar_ns:
            yield None, (item_x, item_y, corr_sim, cos_sim, reg_corr_sim,
                         jaccard_sim, n)


def correlation(size, dot_product, rating_sum, \
            rating2sum, rating_norm_squared, rating2_norm_squared):
    '''
The correlation between two vectors A, B is
[n * dotProduct(A, B) - sum(A) * sum(B)] /
sqrt{ [n * norm(A)^2 - sum(A)^2] [n * norm(B)^2 - sum(B)^2] }

'''
    numerator = size * dot_product - rating_sum * rating2sum
    denominator = sqrt(size * rating_norm_squared - rating_sum * rating_sum) * \
                    sqrt(size * rating2_norm_squared - rating2sum * rating2sum)

    return (numerator / (float(denominator))) if denominator else 0.0


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


def normalized_correlation(size, dot_product, rating_sum, \
            rating2sum, rating_norm_squared, rating2_norm_squared):
    '''
The correlation between two vectors A, B is
cov(A, B) / (stdDev(A) * stdDev(B))
The normalization is to give the scale between [0,1].

'''
    similarity = correlation(size, dot_product, rating_sum, \
            rating2sum, rating_norm_squared, rating2_norm_squared)

    return (similarity + 1.0) / 2.0


def cosine(dot_product, rating_norm_squared, rating2_norm_squared):
    '''
The cosine between two vectors A, B
dotProduct(A, B) / (norm(A) * norm(B))
'''
    numerator = dot_product
    denominator = rating_norm_squared * rating2_norm_squared

    return (numerator / (float(denominator))) if denominator else 0.0


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, \
            rating2sum, rating_norm_squared, rating2_norm_squared)

    w = size / float(size + virtual_cont)

    return w * unregularizedCorrelation + (1.0 - w) * prior_correlation
Something went wrong with that request. Please try again.