Building a Recommendation Engine in Python

Recommendation Engines in Python

import networkx
from operator import itemgetter
import matplotlib.pyplot

# read the data from the amazon-books.txt;
# populate amazonProducts nested dicitonary;
# key = ASIN; value = MetaData associated with ASIN
fhr = open('./amazon-books.txt', 'r', encoding='utf-8', errors='ignore')
amazonBooks = {}
fhr.readline()
for line in fhr:
    cell = line.split('\t')
    MetaData = {}
    MetaData['Id'] = cell[0].strip() 
    ASIN = cell[1].strip()
    MetaData['Title'] = cell[2].strip()
    MetaData['Categories'] = cell[3].strip()
    MetaData['Group'] = cell[4].strip()
    MetaData['Copurchased'] = cell[5].strip()
    MetaData['SalesRank'] = int(cell[6].strip())
    MetaData['TotalReviews'] = int(cell[7].strip())
    MetaData['AvgRating'] = float(cell[8].strip())
    MetaData['DegreeCentrality'] = int(cell[9].strip())
    MetaData['ClusteringCoeff'] = float(cell[10].strip())
    amazonBooks[ASIN] = MetaData
fhr.close()

# read the data from amazon-books-copurchase.adjlist;
# assign it to copurchaseGraph weighted Graph;
# node = ASIN, edge= copurchase, edge weight = category similarity
fhr=open("amazon-books-copurchase.edgelist", 'rb')
copurchaseGraph=networkx.read_weighted_edgelist(fhr)
fhr.close()

# now let's assume a person is considering buying the following book;
# what else can we recommend to them based on copurchase behavior 
# we've seen from other users?
print ("Looking for Recommendations for Customer Purchasing this Book:")
print ("--------------------------------------------------------------")
asin = '0805047905'

# Understand the metadata associated with this book by printing out the
# features associates with the book
print ("ASIN = ", asin) 
print ("Title = ", amazonBooks[asin]['Title'])
print ("SalesRank = ", amazonBooks[asin]['SalesRank'])
print ("TotalReviews = ", amazonBooks[asin]['TotalReviews'])
print ("AvgRating = ", amazonBooks[asin]['AvgRating'])
print ("DegreeCentrality = ", amazonBooks[asin]['DegreeCentrality'])
print ("ClusteringCoeff = ", amazonBooks[asin]['ClusteringCoeff'])
print()


# Create variable dcl from the copurchaseGraph data using the networkx.degree package 
# Create new variable dc is equal to dcl of given asin 
# print dc
dcl = networkx.degree(copurchaseGraph)
dc = dcl[asin]
print ("Degree Centrality:", dc)
print()

# Get ego network of given asin at depth 1 using networkx.ego_graph package
# and assign to variable ego
# print number of nodes in ego 
# print number of edges in ego

ego = networkx.ego_graph(copurchaseGraph, asin, radius=1)
print ("Ego Network:", 
       "Nodes =", ego.number_of_nodes(), 
        "Edges =", ego.number_of_edges())
print()

# Get clustering coefficient of given asin
# Get clustering coefficient of ego using networkx.average_clustering and 
# assign to variable cc
# print clustering coefficient, round to two decimal places

cc = networkx.average_clustering(ego)
print ("Clustering Coefficient:", round(cc,2))
print()

# Use island method on ego network with a threshold of 0.65 to trim down the 
# ego network
# Set threshold to 0.68
# Create empty tuple called egotrim using the networkx.Graph() to represent 
# the trimmed network
# loop node 1, node 2, edge in the ego network edges data:
#    if edge weight is greater than or equal to the threshold:
#        add node 1, node 2, edge weight to the egotrim tuple
#print threshold of the trimmed network
#print number of nodes in the egotrim tuple 
#print number of edges in the egotrim tuple
# print list of egotrim network to obtain the asin of the books in the 
# trimmed network

threshold = 0.68
egotrim = networkx.Graph()
for n1, n2, e in ego.edges(data=True):
    if e['weight'] >= threshold:
        egotrim.add_edge(n1,n2,e)
print ("Trimmed Ego Network:", 
       "Threshold=", threshold,
       "Nodes =", egotrim.number_of_nodes(), 
        "Edges =", egotrim.number_of_edges())
print()

print("Asin in the trimmed network: ", list(egotrim))

# Print out the neighbors of the asin in the trimmed network
# Create variable called neighbors which contains the neighbors of the 
# given asin in the trimmed network 
# print list of neighbors

print()
neighbors = egotrim.neighbors(asin)
print()

print("Neighbours in the trimmed network: ", list(neighbors))

# Write a for loop statement to reiterate in the neighbors tuple and 
# print out the recommendations for customers 
# purchasing this book
# Loop with neighbor asin as nb_asin in neighbors tuple:
#    Print nb_asin 
#    Print Title from the amazonBooks dataset
#    Print AvgRating from the amazonBooks dataset
#    Print TotalReviews from the amazonBooks dataset

print()
for nb_asin in neighbors:
    print("Asin: ", nb_asin)
    print("Book Title: ", amazonBooks[nb_asin]["Title"])
    print("Average Rating:", amazonBooks[nb_asin]["AvgRating"])
    print("Number of Reviews: ", amazonBooks[nb_asin]["TotalReviews"])