def main():
timeStart = time()
print('Pre-processing cf data...')
train_data = readFile('./data/train.csv', separator=' ', columns=['userID', 'movieID', 'rating'],
types={'userID': np.int32, 'movieID': np.int32, 'rating': np.float32})
train_data = train_data.sample(n=len(train_data))
test_data = readFile('./data/test.csv', separator=' ', columns=['userID', 'movieID'],
types={'userID': np.int32, 'movieID': np.int32})
reader = Reader(rating_scale=(0, 5))
train_data = Dataset.load_from_df(train_data, reader)
model = SVD(n_factors=5, n_epochs=50)
train_data = train_data.build_full_trainset()
cf_predictions = cf_predict(train_data, test_data, model)
print('\ncf train-predict costs (%d seconds)' % (time() - timeStart))
with open("./data/train.csv", "r") as data:
data = data.readlines()
for row in data[1:]:
tokens = word_tokenize(row)
if (user_movie_rating.get(int(tokens[0])) == None):
user_movie_rating[int(tokens[0])] = []
user_movie_rating[int(tokens[0])].append((int(tokens[1]), float(tokens[2])))
fill_content_dict("./data/additional_files/movie_genres.data", movie_genres)
fill_content_dict("./data/additional_files/movie_directors.data", movie_directors)
fill_content_dict("./data/additional_files/movie_actors.data", movie_actors)
fill_content_dict("./data/additional_files/movie_tags.data", movie_tags)
print("Done! (%d seconds)\n" % (time() - timeStart))
print("Predicting...")
timePred = time()
content_predictions = []
with open("./data/test.csv", "r") as testFile:
for row in testFile.readlines()[1:]:
tokens = word_tokenize(row)
tmp = contentBased_predict(int(tokens[0]), int(tokens[1]))
content_predictions.append(contentBased_predict(int(tokens[0]), int(tokens[1])))
print("Done! (%d seconds)\n" % (time() - timePred))
print("Writing to file...")
with open("./data/res.data", 'w') as predFile:
for p1, p2 in zip(content_predictions, cf_predictions):
predFile.write("%f\n" % ((float(p1)+float(p2))/2))
def main():
timeStart = time()
#preprocess data
print('Pre-processing cf data...')
# Read train (shuffled) and test data as DataFrames
train_data = readFile('./data/train.csv',
separator=' ',
columns=['userID', 'movieID', 'rating'],
types={
'userID': np.int32,
'movieID': np.int32,
'rating': np.float32
})
train_data = train_data.sample(n=len(train_data))
test_data = readFile('./data/test.csv',
separator=' ',
columns=['userID', 'movieID'],
types={
'userID': np.int32,
'movieID': np.int32
})
# Build the train data as a Surprise's DataSet object
reader = Reader(rating_scale=(0, 5)) # Standardized rating scale
train_data = Dataset.load_from_df(train_data, reader)
model = SVD(n_factors=5, n_epochs=50)
# Build a Trainset object to feed into the prediction algorithm.
train_data = train_data.build_full_trainset()
# Predict ratings for each user and associated movie
cf_predictions = cf_predict(train_data, test_data, model)
print('\ncf train-predict costs (%d seconds)' % (time() - timeStart))
#create dict {userID: [(movieID1, rating1), (movieID2, rating2), ...]}
#only includes ratings for movies that the user has seen
with open("./data/train.csv", "r") as data:
data = data.readlines()
for row in data[1:]:
tokens = word_tokenize(row)
if (user_movie_rating.get(int(tokens[0])) == None):
user_movie_rating[int(tokens[0])] = []
user_movie_rating[int(tokens[0])].append(
(int(tokens[1]), float(tokens[2])))
#create dict {movieID: [genre1, genre2, ...]}
fill_content_dict("./data/additional_files/movie_genres.data",
movie_genres)
#create dict {movieID: director}
fill_content_dict("./data/additional_files/movie_directors.data",
movie_directors)
#create dict {movieID: [actor1, actor2, ...]}
fill_content_dict("./data/additional_files/movie_actors.data",
movie_actors)
#create dict {movieID: [tag1, tag2, ...]}
fill_content_dict("./data/additional_files/movie_tags.data", movie_tags)
print("Done! (%d seconds)\n" % (time() - timeStart))
#predict ratings
print("Predicting...")
timePred = time()
content_predictions = []
with open("./data/test.csv", "r") as testFile:
for row in testFile.readlines()[1:]:
tokens = word_tokenize(row)
content_predictions.append(
contentBased_predict(int(tokens[0]), int(tokens[1])))
print("Done! (%d seconds)\n" % (time() - timePred))
#write results to file
print("Writing to file...")
timeWrite = time()
with open("./data/res.data", 'w') as predFile:
for p1, p2 in zip(content_predictions, cf_predictions):
predFile.write("%f\n" % ((float(p1) + float(p2)) / 2))
