def predict(self,
user,
books,
ratings,
prediction_num,
already_read=None,
graph=None):
# Predicts top ten unread books for a given user ID
if graph is None:
if already_read is None:
already_read = ratings[ratings['user_id'] ==
user]['book_id'].unique()
book_data = np.array(
list(set(books[~books['book_id'].isin(already_read)].book_id)))
user_data = np.array([user for i in range(len(book_data))])
predictions = self.model.predict([user_data, book_data],
batch_size=512)
predictions = np.array([item[0] for item in predictions])
recommended_book_ids = (-predictions).argsort()
recommended_books = books[books['book_id'].isin(
recommended_book_ids)]
result = []
for _, book in recommended_books.iterrows():
b = Book(book['book_id'], book['title'], book['authors'],
book['average_rating'], book['image_url'])
result.append(b)
return result
else:
with graph.as_default():
if already_read is None:
already_read = ratings[ratings['user_id'] ==
user]['book_id'].unique()
book_data = np.array(
list(
set(books[~books['book_id'].isin(already_read)].book_id
)))
user_data = np.array([user for i in range(len(book_data))])
predictions = self.model.predict([user_data, book_data],
batch_size=512)
predictions = np.array([item[0] for item in predictions])
recommended_book_ids = (
-predictions).argsort()[:prediction_num]
recommended_books = books[books['book_id'].isin(
recommended_book_ids)]
result = []
for _, book in recommended_books.iterrows():
b = Book(book['book_id'], book['title'], book['authors'],
book['average_rating'], book['image_url'])
result.append(b)
return result
def convert(recommendations):
result = []
for _, book in recommendations.iterrows():
b = Book(book['book_id'], book['title'], book['authors'],
book['average_rating'], book['image_url'])
result.append(b)
return result
def search(search_param, ):
global results, books
found_books = books[books['title'].str.contains(search_param)][:10]
results = []
for _, row in found_books.iterrows():
book = Book(row['book_id'], row['title'], row['authors'],
row['average_rating'], row['image_url'])
results.append(book)
return results
def neumf_predict(user_id=None):
global results, ncf, selected, graph
if user_id is not None:
user_id = int(user_id)
rated = ratings[ratings['user_id'] == user_id]['book_id'].unique()
read = books[books['book_id'].isin(rated)][:num_predict]
already_read = []
for _, book in read.iterrows():
b = Book(book['book_id'], book['title'], book['authors'],
book['average_rating'], book['image_url'])
already_read.append(b)
results = ncf.predict(user_id, books, ratings, num_predict, None,
graph)
return render_template('index.html',
result=results,
selected=already_read)
else:
results = ncf.predict(0, books, ratings, num_predict, selected, graph)
return render_template('index.html',
result=results,
selected=selected_titles)