Books Recommender system using clustering and Collaborative filtering
All about book recommendation system.
- Creating Environment for the project
- Preprocessing the book dataset
- Preprocessing user csv file
- Preprocessing bookrating csv file
- Now Model Training Begins
- Finding url
- Testing model
import pandas as pd
import numpy as np
import pickle
from scipy.sparse import csr_matrix
from sklearn.neighbors import NearestNeighbors
books = pd.read_csv('data/books.csv', sep=";", error_bad_lines=False, encoding='latin-1')
books.sample(2)
books.iloc[237]['Image-URL-L']
books.shape
books.columns
- Here we are selecting the most important columns and also selecting Image URL columns for the poster.
books = books[['ISBN','Book-Title', 'Book-Author', 'Year-Of-Publication', 'Publisher','Image-URL-L']]
books.sample(5)
books.rename(columns={"Book-Title":'title',
'Book-Author':'author',
"Year-Of-Publication":'year',
"Publisher":"publisher",
"Image-URL-L":"image_url"},inplace=True)
books.sample(5)
users = pd.read_csv('./data/user.csv', sep=";", error_bad_lines=False, encoding='latin-1')
users.sample(5)
users.shape
users.rename(columns={"User-ID":'user_id',
'Location':'location',
"Age":'age'},inplace=True)
users.sample(5)
ratings = pd.read_csv('./data/bookrating.csv', sep=";", error_bad_lines=False, encoding='latin-1')
ratings.sample(5)
ratings.shape
ratings.rename(columns={"User-ID":'user_id',
'Book-Rating':'rating'},inplace=True)
ratings.sample(5)
- Now we have 3 dataframes
booksusersratings
print(books.shape, users.shape, ratings.shape, sep='\n')
ratings['user_id'].value_counts()
ratings['user_id'].value_counts().shape
ratings['user_id'].unique().shape
x = ratings['user_id'].value_counts() > 200
x[x].shape
y= x[x].index
y
ratings = ratings[ratings['user_id'].isin(y)]
ratings.head()
ratings.shape
ratings_with_books = ratings.merge(books, on='ISBN')
ratings_with_books.head()
ratings_with_books.shape
number_rating = ratings_with_books.groupby('title')['rating'].count().reset_index()
number_rating.sample(5)
number_rating.rename(columns={'rating':'num_of_rating'},inplace=True)
number_rating.head()
final_rating = ratings_with_books.merge(number_rating, on='title')
final_rating.head()
final_rating.shape
final_rating = final_rating[final_rating['num_of_rating'] >= 50]
final_rating.head()
final_rating.shape
final_rating.drop_duplicates(['user_id','title'],inplace=True)
final_rating.shape
book_pivot = final_rating.pivot_table(columns='user_id', index='title', values= 'rating')
book_pivot
book_pivot.shape
book_pivot.fillna(0, inplace=True)
book_pivot
- We are using scipy csr_matrix for fast row slicing, faster matrix vector products.
book_sparse = csr_matrix(book_pivot)
- Now import our clustering algoritm which is Nearest Neighbors this is an unsupervised ml algorithm and used in our model building part.
model = NearestNeighbors(algorithm= 'brute')
model.fit(book_sparse)
- We are calculating distance and suggestion using k nearest neighbours.
distance, suggestion = model.kneighbors(book_pivot.iloc[237,:].values.reshape(1,-1), n_neighbors=6 )
pickle.dump(model, open('./model/model.pkl','wb'))
distance
suggestion
book_pivot.iloc[241,:]
- Now printing a suggested book by K-Nearest Neighbour.
for i in range(len(suggestion)):
print(book_pivot.index[suggestion[i]])
book_pivot.index[5]
book_names = book_pivot.index
book_names[2]
np.where(book_pivot.index == 'A Bend in the Road')[0][0]
ids = np.where(final_rating['title'] == "Harry Potter and the Chamber of Secrets (Book 2)")[0][0]
final_rating.iloc[ids]['image_url']
- Storing book name by the id given by suggestion in knearest neaghbours.
book_name = []
for book_id in suggestion:
book_name.append(book_pivot.index[book_id])
book_name[0]
- Extracting index of id from a specific book name.
ids_index = []
for name in book_name[0]:
ids = np.where(final_rating['title'] == name)[0][0]
ids_index.append(ids)
- Now extracting book given by ids_index
for idx in ids_index:
url = final_rating.iloc[idx]['image_url']
print(url)
- Now storing model and other related file in pickle format for prediction purpose.
pickle.dump(model,open('artifacts/model.pkl','wb'))
pickle.dump(book_names,open('artifacts/book_names.pkl','wb'))
pickle.dump(final_rating,open('artifacts/final_rating.pkl','wb'))
pickle.dump(book_pivot,open('artifacts/book_pivot.pkl','wb'))
def recommend_book(book_name):
book_id = np.where(book_pivot.index == book_name)[0][0]
distance, suggestion = model.kneighbors(book_pivot.iloc[book_id,:].values.reshape(1,-1), n_neighbors=6 )
for i in range(len(suggestion)):
books = book_pivot.index[suggestion[i]]
for j in books:
if j == book_name:
print(f"You searched '{book_name}'\n")
print("The suggestion books are: \n")
else:
print(j)
book_name = "Harry Potter and the Chamber of Secrets (Book 2)"
recommend_book(book_name)