victorkohler

import pandas as pd
import numpy as np

from sklearn.metrics.pairwise import cosine_similarity
from scipy import sparse

#------------------
# LOAD THE DATASET
#------------------

victorkohler

#------------------------
# ITEM-ITEM CALCULATIONS
#------------------------

# As a first step we normalize the user vectors to unit vectors.

# magnitude = sqrt(x2 + y2 + z2 + ...)
magnitude = np.sqrt(np.square(data_items).sum(axis=1))

victorkohler

#------------------------
# USER-ITEM CALCULATIONS
#------------------------

user = 5985 # The id of the user for whom we want to generate recommendations
user_index = data[data.user == user].index.tolist()[0] # Get the frame index

# Get the artists the user has likd.
known_user_likes = data_items.ix[user_index]

victorkohler

#------------------------
# USER-ITEM CALCULATIONS
#------------------------

# Construct a new dataframe with the 10 closest neighbours (most similar)
# for each artist.
data_neighbours = pd.DataFrame(index=data_matrix.columns, columns=range(1,11))
for i in xrange(0, len(data_matrix.columns)):
    data_neighbours.ix[i,:10] = data_matrix.ix[0:,i].sort_values(ascending=False)[:10].index

victorkohler

import pandas as pd
import numpy as np

from sklearn.metrics.pairwise import cosine_similarity
from scipy import sparse

#------------------
# LOAD THE DATASET
#------------------

victorkohler

user,a perfect circle,abba,ac/dc,adam green,aerosmith,afi,air,alanis morissette,alexisonfire,alicia keys,all that remains,amon amarth,amy macdonald,amy winehouse,anti-flag,aphex twin,apocalyptica,arcade fire,arch enemy,arctic monkeys,as i lay dying,atb,atreyu,audioslave,avril lavigne,babyshambles,bad religion,beastie boys,beatsteaks,beck,beirut,belle and sebastian,beyonce,billy talent,bjork,black eyed peas,black sabbath,blind guardian,blink-182,bloc party,bloodhound gang,blur,boards of canada,bob dylan,bob marley,bob marley & the wailers,breaking benjamin,bright eyes,bring me the horizon,britney spears,bruce springsteen,bullet for my valentine,caliban,cascada,cat power,children of bodom,chimaira,christina aguilera,clueso,cocorosie,coldplay,crystal castles,cypress hill,daft punk,damien rice,dark tranquillity,david bowie,david guetta,death cab for cutie,deep purple,deftones,deichkind,depeche mode,dido,die apokalyptischen reiter,die toten hosen,digitalism,dimmu borgir,dire straits,disturbed,dream theater,dredg,d

victorkohler

import random
import pandas as pd
import numpy as np

import scipy.sparse as sparse
from scipy.sparse.linalg import spsolve
from sklearn.preprocessing import MinMaxScaler

#-------------------------
# LOAD AND PREP THE DATA

victorkohler

def implicit_als(sparse_data, alpha_val=40, iterations=10, lambda_val=0.1, features=10):
 
""" Implementation of Alternating Least Squares with implicit data. We iteratively
    compute the user (x_u) and item (y_i) vectors using the following formulas:
 
    x_u = ((Y.T*Y + Y.T*(Cu - I) * Y) + lambda*I)^-1 * (X.T * Cu * p(u))
    y_i = ((X.T*X + X.T*(Ci - I) * X) + lambda*I)^-1 * (Y.T * Ci * p(i))
 
    Args:

victorkohler

""" Continuation of implicit_als function"""

    # Start main loop. For each iteration we first compute X and then Y
    for i in xrange(iterations):
        print 'iteration %d of %d' % (i+1, iterations)
        
        # Precompute Y-transpose-Y and X-transpose-X
        yTy = Y.T.dot(Y)
        xTx = X.T.dot(X)

victorkohler

user_vecs, item_vecs = implicit_als(data_sparse, iterations=20, features=20, alpha_val=40)