def get_sim_matrix(load_sim_matrix):
raw = pd.read_csv('books_metadata/ratings.csv')
if load_sim_matrix is False:
raw = pd.read_csv('books_metadata/ratings.csv')
raw = raw[raw['book_id'] <= 1900]
raw.drop_duplicates(inplace=True)
print('we have', str(raw.shape[0]), 'ratings')
print('the number of unique users we have is:',
len(raw.user_id.unique()))
print('the number of unique books we have is:',
len(raw.book_id.unique()))
rawTrain = raw[['user_id', 'book_id', 'rating']]
rawTrain, rawHoldout = train_test_split(raw, test_size=0.25)
reader = surprise.Reader(rating_scale=(1, 5))
data = surprise.Dataset.load_from_df(rawTrain, reader)
sim_options = {'name': 'cosine', 'user_based': False}
collabKNN = surprise.KNNWithMeans(k=100, sim_options=sim_options)
kSplit = surprise.model_selection.split.KFold(n_splits=2,
shuffle=False)
for trainset, testset in kSplit.split(data):
collabKNN.fit(trainset)
predictionsKNN = collabKNN.test(testset)
surprise.accuracy.rmse(predictionsKNN, verbose=True)
sim_matrix = collabKNN.compute_similarities()
with open('books_sim_matrix', 'wb') as output:
pickle.dump(sim_matrix, output, protocol=pickle.HIGHEST_PROTOCOL)
return sim_matrix
with open('books_sim_matrix', 'rb') as input:
sim_matrix = pickle.load(input)
return sim_matrix
def train(ratings, k_neighbors, k_folds):
"""
Train a model and return it. Then we can use the model and evaluate it elsewhere
@param ratings dataframe pandas dataframe to train on, with columns UserId, MovieId, Ratings
@param k_neighbors number of neighbors to examine
@param k_folds number of folds for cross validation
@returns List of (algo, test data)
We can call methods such as `test` and `evaluate` on this object
"""
train_data, test_data = cv.train_test_split(ratings, test_size=0.20)
reader = sp.Reader(rating_scale=(1, 5))
trainset = sp.Dataset.load_from_df(train_data, reader)
testset = sp.Dataset.load_from_df(test_data, reader)
trainset.split(n_folds=k_folds)
similarity_options = {'name': 'pearson', 'user_based': False}
algo = sp.KNNWithMeans(sim_options=similarity_options,
k=k_neighbors,
min_k=5)
for _trainset, _ in trainset.folds():
algo.train(_trainset)
testset = testset.build_full_trainset().build_testset()
return (algo, testset)
def meansKNN(train, test):
"""
Run the KNN means model from Surprise library.
@param train: the training set in the Surprise format.
@param test: the test set in the Surprise format.
@return: the predictions in a numpy array.
"""
algo = spr.KNNWithMeans()
algo.fit(train)
predictions = algo.test(test)
return get_predictions(predictions)
def algo_tester(data_object):
'''
Produces a dataframe displaying all the different RMSE's, test & train times of the different surprise algorithms
---Parameters---
data_object(variable) created from the read_data_surprise function
---Returns---
returns a dataframe where you can compare the performance of different algorithms
'''
benchmark = []
algos = [
sp.SVDpp(),
sp.SVD(),
sp.SlopeOne(),
sp.NMF(),
sp.NormalPredictor(),
sp.KNNBaseline(),
sp.KNNBasic(),
sp.KNNWithMeans(),
sp.KNNWithZScore(),
sp.BaselineOnly(),
sp.CoClustering()
]
# Iterate over all algorithms
for algorithm in algos:
# Perform cross validation
results = cross_validate(algorithm,
data_object,
measures=['RMSE'],
cv=3,
verbose=False)
# Get results & append algorithm name
tmp = pd.DataFrame.from_dict(results).mean(axis=0)
tmp = tmp.append(
pd.Series([str(algorithm).split(' ')[0].split('.')[-1]],
index=['Algorithm']))
benchmark.append(tmp)
benchmark = pd.DataFrame(benchmark).set_index('Algorithm').sort_values(
'test_rmse')
return benchmark
uid_list = [2, 3, 4, 5, 6, 7, 8, 9, 10, 11]
# TODO - 2-1-1. KNNBasic, cosine
sim_options = {'name': 'cosine'}
algo = surprise.KNNBasic(sim_options=sim_options)
algo.fit(trainset)
results = get_top_n(algo, testset, uid_list, n=10, user_based=True)
with open('2-1-1_results.txt', 'w') as f:
for uid, ratings in sorted(results.items(), key=lambda x: int(x[0])):
f.write('User ID %s top-10 results\n' % uid)
for iid, score in ratings:
f.write('Item ID %s\tscore %s\n' % (iid, str(score)))
f.write('\n')
# TODO - 2-1-2. KNNWithMeans, pearson
sim_options2 = {'name': 'pearson'}
algo = surprise.KNNWithMeans(sim_options=sim_options2)
algo.fit(trainset)
results = get_top_n(algo, testset, uid_list, n=10, user_based=True)
with open('2-1-2_results.txt', 'w') as f:
for uid, ratings in sorted(results.items(), key=lambda x: int(x[0])):
f.write('User ID %s top-10 results\n' % uid)
for iid, score in ratings:
f.write('Item ID %s\tscore %s\n' % (iid, str(score)))
f.write('\n')
# # TODO - 2-2. Best Model
bsl_options_ub = {'n_epochs': 30, 'method': 'als', 'reg_i': 10, 'reg_u': 0}
sim_options_ub = {'name': 'msd', 'min_support': 1, 'user_based': True}
best_algo_ub = surprise.KNNBaseline(sim_options=sim_options_ub,
bsl_options=bsl_options_ub,
def main(args):
parser = argparse.ArgumentParser(description= \
'Deploys recommendation algorithms and outputs the recommendations list',\
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument("--pickleLoadPath", type=str, action='store', \
help= 'If set=> load topN recoms from pickle file')
parser.add_argument("--pickleSavePath",
type=str,
action='store',
help='If set => Output .pickle file.')
parser.add_argument("--proc", type=int, default=multiprocessing.cpu_count(), \
action='store', \
help= 'Number of processes to spawn for topN computation\n' +
'default is number of processors.')
parser.add_argument("--update_freq", type=int, default=1, action='store', \
help= 'Number of clicks after which the model is updated')
parser.add_argument("--topN_list", type=int, nargs="+", required=True, \
help= 'e.g., --topN_list 5 10 50\n' \
+ 'topN=max(topN_list); the rest of the values are used for evaluation.')
parser.add_argument("--drop_ratio", type=int, default=0, action='store', \
help= 'Number of random events to remove from the training set;\n' + \
'default is 0; Currently not implemented for librec.')
parser.add_argument("--evalTrain", dest='evalTrain', action='store_true', \
help='If set => evaluate on training set using k-fold validation.\n' \
+ 'Else => evaluate only on test set')
parser.add_argument("--dataset", type=str, action='store', \
help= 'Full path to the dataset.\n' + \
'Must give --testSize and --validSize for the split')
parser.add_argument("--testSize",
type=int,
default=0,
action='store',
help='TestSet size; default is 0 => no test set')
parser.add_argument("--validSize", type=int, default=2000, action='store', \
help= 'Validation Set size; default is 2000.')
parser.add_argument("--trainSet", type=str, action='store', \
help= 'Full path to the trainingSet.csv\n' + \
'If given the (potential) training set split from --dataset will be overwritten')
parser.add_argument("--validSet", type=str, action='store', \
help= 'Full path to the validationSet.csv\n' + \
'If given the (potential) validation set split from --dataset will be overwritten')
parser.add_argument("--testSet", type=str, action='store', \
help= 'Full path to the testSet.csv\n' + \
'If given the (potential) test set split from --dataset will be overwritten')
parser.add_argument("--librec_home", type=str, action='store', \
help= 'Full path to the librec folder cloned from git.')
parser.add_argument("--config", type=str, action='store', \
help= 'Full path to the librec .properties file.\n' + \
'Copy from: https://www.librec.net/dokuwiki/doku.php?id=AlgorithmList')
parser.add_argument("--surprise_algo", type=str, action='store', \
help= 'Choose algorithm from surprise lib. Available options:\n' + \
'--surprise_algo SVD\n' + \
'--surprise_algo SVDpp\n' + \
'--surprise_algo PMF\n' + \
'--surprise_algo NMF\n' + \
'--surprise_algo KNNWithMeans\n')
args = parser.parse_args(args)
random.seed(42) # reproducability
np.random.seed(42)
if args.pickleLoadPath is None:
"""DATA"""
train, valid, test = splitter.splitData(
fullDataPath=args.dataset, validSize=args.validSize, testSize=args.testSize, \
trainSetPath=args.trainSet, validSetPath=args.validSet, testSetPath=args.testSet)
"""RECOMMENDATIONS"""
if args.surprise_algo == 'SVD':
algo = surprise.SVD()
elif args.surprise_algo == 'KNNWithMeans':
# sim_options = {'name': 'pearson_baseline', 'shrinkage': 2500, \
# 'user_based': False, }
sim_options = {'name': 'cosine', 'user_based': False}
algo = surprise.KNNWithMeans(k=40, sim_options=sim_options)
elif args.surprise_algo == 'PMF':
algo = surprise.SVD(n_factors=5,
reg_all=0.12,
lr_all=0.005,
n_epochs=400)
elif args.surprise_algo == 'NMF':
algo = surprise.NMF(n_factors=5, n_epochs=400)
elif args.surprise_algo == 'SVDpp':
algo = surprise.SVDpp()
testList = [] # output recommendations for the last element
if len(test) > 0:
testList.append(test)
if len(valid) > 0:
testList.append(valid)
for test in testList:
if args.librec_home is None:
recs = surprise_recom(train, test, algo, drop_ratio=args.drop_ratio, \
update_freq=args.update_freq, N_list=args.topN_list, num=args.proc, \
evalTrain=args.evalTrain)
else:
recs = librec_recom(train, test, args.librec_home, args.config, \
update_freq=args.update_freq, N_list=args.topN_list, num=args.proc, \
evalTrain=args.evalTrain)
if not args.pickleSavePath is None:
with open(args.pickleSavePath, 'wb') as handle:
pickle.dump(recs, handle)
else:
with open(args.pickleLoadPath, 'rb') as handle:
recs = pickle.load(handle)
from surprise import model_selection
import pandas as pd
import csv
def read_train_data(path):
file_path = os.path.normpath(path)
reader = surprise.Reader(line_format='timestamp user item rating', sep=',')
data = surprise.Dataset.load_from_file(file_path, reader=reader)
return data
movie_train = read_train_data('E:/train_v2.csv')
print(movie_train.raw_ratings)
knn_estimator = surprise.KNNWithMeans()
knn_grid = {
'k': [10, 20],
'sim_options': {
'name': ['cosine'],
'min_support': [1, 5],
'user_based': [False]
}
}
knn_grid_estimator = model_selection.GridSearchCV(knn_estimator,
knn_grid,
measures=['rmse'],
cv=3)
#do grid search using cv strategy
knn_grid_estimator.fit(movie_train)
print(knn_grid_estimator.best_score['rmse'])
def main(train_df, target_df, cache_name="test", force_recompute=[]):
"""Train multiple models on train_df and predicts target_df
Predictions are cached. If the indices don't match the indices of
target_df, the cache is discarded.
By default, if a method was already computed it is not recomputed again
(except if the method name is listed in force_recompute). cache_name
is the name to use to read and write the cache.
Arguments:
train_df {dataframe} -- Training dataframe
target_df {dataframe} -- Testing dataframe
Keyword Arguments:
cache_name {str} -- Name to use for caching (default: {"test"})
force_recompute {list} -- Name(s) of methods to recompute, whether or
not it was already computed. Useful to only recompute single methods
without discarding the rest. (default: {[]})
Returns:
Dataframe -- Dataframe with predictions for each methods as columns,
IDs as indices
"""
global algo_in_use
CACHED_DF_FILENAME = os.path.dirname(
os.path.abspath(__file__)) +\
"/cache/cached_predictions_{}.pkl".format(cache_name)
train_df = preprocess_df(train_df)
trainset = pandas_to_data(train_df)
ids_to_predict = target_df["Id"].to_list()
# try to retrieve backup dataframe
try:
print("Retrieving cached predictions")
all_algos_preds_df = pd.read_pickle(CACHED_DF_FILENAME)
print("Ensuring cached IDs match given IDs")
assert sorted(ids_to_predict) == sorted(
all_algos_preds_df.index.values)
print("Indices match, continuing")
except (FileNotFoundError, AssertionError):
print("No valid cached predictions found")
all_algos_preds_df = pd.DataFrame(ids_to_predict, columns=["Id"])
all_algos_preds_df.set_index("Id", inplace=True)
all_algos = {
"SVD": spr.SVD(n_factors=200, n_epochs=100),
"Baseline": spr.BaselineOnly(),
"NMF": spr.NMF(n_factors=30, n_epochs=100),
"Slope One": spr.SlopeOne(),
"KNN Basic": spr.KNNBasic(k=60),
"KNN Means": spr.KNNWithMeans(k=60),
"KNN Baseline": spr.KNNBaseline(),
"KNN Zscore": spr.KNNWithZScore(k=60),
"SVD ++": spr.SVDpp(n_factors=40, n_epochs=100),
"Co Clustering": spr.CoClustering()
}
for name in all_algos:
print("##### {} ####".format(name))
if name in force_recompute and name in all_algos_preds_df.columns:
all_algos_preds_df.drop(name, axis=1, inplace=True)
if name in all_algos_preds_df.columns:
print("Already computed {}, skipping".format(name))
continue
algo = all_algos[name]
time.sleep(1)
algo.fit(trainset)
time.sleep(1)
algo_in_use = algo
print("Generating predictions...")
predictions = parallelize_predictions(ids_to_predict, 80)
print("Done. Merging with previous results")
this_algo_preds_df = pd.DataFrame(predictions, columns=["Id", name])
this_algo_preds_df.set_index("Id", inplace=True)
all_algos_preds_df = pd.merge(all_algos_preds_df,
this_algo_preds_df,
left_index=True,
right_index=True)
all_algos_preds_df.to_pickle(CACHED_DF_FILENAME)
print("DONE computing surprize")
return all_algos_preds_df
train_path = path + '/Data/train_format.txt'
train_reader = Reader(line_format='user item rating timestamp',
sep=',',
rating_scale=(0, 5))
trainset = Dataset.load_from_file(train_path, reader=train_reader)
trainset = trainset.build_full_trainset()
if args.model == 'NormalPredictor':
model = surprise.NormalPredictor()
elif args.model == 'BaselineOnly':
model = surprise.BaselineOnly()
elif args.model == 'KNNBasic':
model = surprise.KNNBasic()
elif args.model == 'KNNWithMeans':
model = surprise.KNNWithMeans()
elif args.model == 'KNNWithZScore':
model = surprise.KNNWithZScore()
elif args.model == 'KNNBaseline':
model = surprise.KNNBaseline()
elif args.model == 'SVD':
model = surprise.SVD()
elif args.model == 'SVDpp':
model = surprise.SVDpp(verbose=True)
elif args.model == 'NMF':
model = surprise.NMF()
elif args.model == 'SlopeOne':
model = surprise.SlopeOne()
elif args.model == 'CoClustering':
model = surprise.CoClustering()
normalPredictor = surprise.NormalPredictor()
normalPredictor_temp = surprise.model_selection.cross_validate(
normalPredictor, rating_data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
print('normalPredictor--------------')
print(normalPredictor_temp)
baselineOnly = surprise.BaselineOnly()
baselineOnly_temp = surprise.model_selection.cross_validate(
baselineOnly, rating_data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
print('baselineOnly-----------------')
print(baselineOnly_temp)
knnBasic = surprise.KNNBasic()
knnBasic_temp = surprise.model_selection.cross_validate(
knnBasic, rating_data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
print('knnBasic-----------------')
print(knnBasic_temp)
knnWithMeans = surprise.KNNWithMeans()
knnWithMeans_temp = surprise.model_selection.cross_validate(
knnWithMeans, rating_data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
print('knnWithMeans-----------------')
print(knnWithMeans_temp)
knnBaseline = surprise.KNNBaseline()
knnBaseline_temp = surprise.model_selection.cross_validate(
knnBaseline, rating_data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
print('knnBaseline-----------------')
print(knnBaseline_temp)
svdpp = surprise.SVDpp()
svdpp_temp = surprise.model_selection.cross_validate(svdpp,
rating_data,
measures=['RMSE', 'MAE'],
cv=5,
verbose=True)
# with open('result.txt', 'w') as file:
# for est_n, true_n in GridList:
# precisions = []
# recalls = []
# for trainset, testset in kf.split(data):
# algo.fit(trainset)
# predictions = algo.test(testset, verbose=False)
# prec, rec = precision_recall(predictions, est_n, true_n)
# precisions.append(prec)
# recalls.append(rec)
# # precisions, recalls = precision_recall_at_k(predictions)
# # # Precision and recall can then be averaged over all users
# # precision.append(sum(prec for prec in precisions.values()) / len(precisions))
# # recall.append(sum(rec for rec in recalls.values()) / len(recalls))
# # mae.append(accuracy.mae(predictions))
# # rmse.append(accuracy.rmse(predictions))
#
# precision = sum(precisions) / 5
# recall = sum(recalls) / 5
# print(precision,recall)
# file.write(str(precision)+'\n')
# file.write(str(recall)+'\n')
# print(sum(mae) / 5)
# print(sum(rmse) / 5)
cross_validate(surprise.BaselineOnly(), data, cv=5, verbose=True)
cross_validate(surprise.KNNBasic(), data, cv=5, verbose=True)
cross_validate(surprise.KNNWithMeans(), data, cv=5, verbose=True)
cross_validate(surprise.KNNBaseline(), data, cv=5, verbose=True)
cross_validate(surprise.SVD(), data, cv=5, verbose=True)
bsl_options = {'method': 'sgd',
'learning_rate': .0005
}
#bsl_options = {'method': 'sgd',
# 'learning_rate': .0005
# }
sim_options = {'name': 'cosine', 'min_support': 5,
'user_based': True,
}
algo1 = surprise.KNNBasic(sim_options=sim_options)
algo2 = surprise.KNNWithMeans(k=15, min_k=5,
sim_options=sim_options)
algo3 = surprise.SVD(n_factors=20000, n_epochs=20, biased=False,
init_mean=0, init_std_dev=0.1, lr_all=0.005,
lr_bu=None, lr_bi=None, lr_qi=None, reg_all=None, reg_bu=0,
reg_bi=0, reg_pu=0, reg_qi=0, verbose=True)
#algo4 = surprise.SVDpp(n_factors=20, n_epochs=20, init_mean=0, init_std_dev=0.1,
# lr_all=0.005, reg_all=None, verbose=True)
print('Training model...')
algo1.train(trainset)
print('Making predictions...')
predictions = algo1.test(test_set)
print('Evaluating results...')
surprise.accuracy.rmse(predictions, verbose=True)
def part3():
file_path = 'DMA_project2_team%02d_part2_UIR.csv' % team
reader = Reader(line_format='user item rating',
sep=',',
rating_scale=(1, 10),
skip_lines=1)
data = Dataset.load_from_file(file_path, reader=reader)
trainset = data.build_full_trainset()
testset = trainset.build_anti_testset()
# TODO: Requirement 3-2. User-based Recommendation
uid_list = [
'ffffbe8d854a4a5a8ab1a381224f5b80', 'ffe2f26d5c174e13b565d026e1d8c503',
'ffdccaff893246519b64d76c3561d8c7', 'ffdb001850984ce69c5f91360ac16e9c',
'ffca7b070c9d41e98eba01d23a920d52'
]
# TODO - set algorithm for 3-2-1
algo = surprise.KNNBasic(k=40,
min_k=1,
sim_options={
'name': 'cosine',
'user_based': True
},
verbose=True)
algo.fit(trainset)
results = get_top_n(algo, testset, uid_list, n=10, user_based=True)
with open('3-2-1.txt', 'w') as f:
for uid, ratings in sorted(results.items(), key=lambda x: x[0]):
f.write('User ID %s top-10 results\n' % uid)
for iid, score in ratings:
f.write('Item ID %s\tscore %s\n' % (iid, str(score)))
f.write('\n')
# TODO - set algorithm for 3-2-2
algo = surprise.KNNWithMeans(k=40,
min_k=1,
sim_options={
'name': 'pearson',
'user_based': True
},
verbose=True)
algo.fit(trainset)
results = get_top_n(algo, testset, uid_list, n=10, user_based=True)
with open('3-2-2.txt', 'w') as f:
for uid, ratings in sorted(results.items(), key=lambda x: x[0]):
f.write('User ID %s top-10 results\n' % uid)
for iid, score in ratings:
f.write('Item ID %s\tscore %s\n' % (iid, str(score)))
f.write('\n')
# TODO - 3-2-3. Best Model
kfold = KFold(n_splits=5, random_state=0)
parameters = {
'k': [30, 40, 50],
'min_k': [1],
'sim_options': {
'name': ['pearson', 'cosine'],
'user_based': [True]
}
}
# Select the best algo with grid search.
print('Grid Search for user based model...')
grid_KNNBasic = GridSearchCV(surprise.KNNBasic,
measures=['rmse'],
param_grid=parameters,
cv=kfold)
grid_KNNWithMeans = GridSearchCV(surprise.KNNWithMeans,
measures=['rmse'],
param_grid=parameters,
cv=kfold)
grid_KNNBasic.fit(data)
grid_KNNWithMeans.fit(data)
best_KNNBasic_score = grid_KNNBasic.best_score['rmse']
best_KNNWithMeans_score = grid_KNNWithMeans.best_score['rmse']
if best_KNNBasic_score < best_KNNWithMeans_score:
algo_name = 'KNNBasic'
best_algo_ub = grid_KNNBasic.best_estimator['rmse']
with_parameters = grid_KNNBasic.best_params['rmse']
score = best_KNNBasic_score
else:
algo_name = 'KNNWithMeans'
best_algo_ub = grid_KNNWithMeans.best_estimator['rmse']
with_parameters = grid_KNNWithMeans.best_params['rmse']
score = best_KNNWithMeans_score
print('The best UB algorithm is', algo_name, 'with', with_parameters,
'\nscore:', score)
# TODO: Requirement 3-3. Item-based Recommendation
iid_list = ['art', 'teaching', 'career', 'college', 'medicine']
# TODO - set algorithm for 3-3-1
algo = surprise.KNNBasic(k=40,
min_k=1,
sim_options={
'name': 'cosine',
'user_based': False
},
verbose=True)
algo.fit(trainset)
results = get_top_n(algo, testset, iid_list, n=10, user_based=False)
with open('3-3-1.txt', 'w') as f:
for iid, ratings in sorted(results.items(), key=lambda x: x[0]):
f.write('Item ID %s top-10 results\n' % iid)
for uid, score in ratings:
f.write('User ID %s\tscore %s\n' % (uid, str(score)))
f.write('\n')
# TODO - set algorithm for 3-3-2
algo = surprise.KNNWithMeans(k=40,
min_k=1,
sim_options={
'name': 'pearson',
'user_based': False
},
verbose=True)
algo.fit(trainset)
results = get_top_n(algo, testset, iid_list, n=10, user_based=False)
with open('3-3-2.txt', 'w') as f:
for iid, ratings in sorted(results.items(), key=lambda x: x[0]):
f.write('Item ID %s top-10 results\n' % iid)
for uid, score in ratings:
f.write('User ID %s\tscore %s\n' % (uid, str(score)))
f.write('\n')
# TODO - 3-3-3. Best Model
kfold = KFold(n_splits=5, random_state=0)
parameters = {
'k': [30, 40, 50],
'min_k': [1],
'sim_options': {
'name': ['pearson', 'cosine'],
'user_based': [False]
}
}
# Select the best algo with grid search.
print('Grid Search for item based model...')
grid_KNNBasic = GridSearchCV(surprise.KNNBasic,
measures=['rmse'],
param_grid=parameters,
cv=kfold)
grid_KNNWithMeans = GridSearchCV(surprise.KNNWithMeans,
measures=['rmse'],
param_grid=parameters,
cv=kfold)
grid_KNNBasic.fit(data)
grid_KNNWithMeans.fit(data)
best_KNNBasic_score = grid_KNNBasic.best_score['rmse']
best_KNNWithMeans_score = grid_KNNWithMeans.best_score['rmse']
if best_KNNBasic_score < best_KNNWithMeans_score:
algo_name = 'KNNBasic'
best_algo_ub = grid_KNNBasic.best_estimator['rmse']
with_parameters = grid_KNNBasic.best_params['rmse']
score = best_KNNBasic_score
else:
algo_name = 'KNNWithMeans'
best_algo_ub = grid_KNNWithMeans.best_estimator['rmse']
with_parameters = grid_KNNWithMeans.best_params['rmse']
score = best_KNNWithMeans_score
print('The best IB algorithm is', algo_name, 'with', with_parameters,
'\nscore:', score)
# TODO: Requirement 3-4. Matrix-factorization Recommendation
# TODO - set algorithm for 3-4-1
algo = surprise.SVD(n_factors=100, n_epochs=50, biased=False)
algo.fit(trainset)
results = get_top_n(algo, testset, uid_list, n=10, user_based=True)
with open('3-4-1.txt', 'w') as f:
for uid, ratings in sorted(results.items(), key=lambda x: x[0]):
f.write('User ID %s top-10 results\n' % uid)
for iid, score in ratings:
f.write('Item ID %s\tscore %s\n' % (iid, str(score)))
f.write('\n')
# TODO - set algorithm for 3-4-2
algo = surprise.SVD(n_factors=200, n_epochs=100, biased=True)
algo.fit(trainset)
results = get_top_n(algo, testset, uid_list, n=10, user_based=True)
with open('3-4-2.txt', 'w') as f:
for uid, ratings in sorted(results.items(), key=lambda x: x[0]):
f.write('User ID %s top-10 results\n' % uid)
for iid, score in ratings:
f.write('Item ID %s\tscore %s\n' % (iid, str(score)))
f.write('\n')
# TODO - set algorithm for 3-4-3
algo = surprise.SVDpp(n_factors=100, n_epochs=50)
algo.fit(trainset)
results = get_top_n(algo, testset, uid_list, n=10, user_based=True)
with open('3-4-3.txt', 'w') as f:
for uid, ratings in sorted(results.items(), key=lambda x: x[0]):
f.write('User ID %s top-10 results\n' % uid)
for iid, score in ratings:
f.write('Item ID %s\tscore %s\n' % (iid, str(score)))
f.write('\n')
# TODO - set algorithm for 3-4-4
algo = surprise.SVDpp(n_factors=100, n_epochs=100)
algo.fit(trainset)
results = get_top_n(algo, testset, uid_list, n=10, user_based=True)
with open('3-4-4.txt', 'w') as f:
for uid, ratings in sorted(results.items(), key=lambda x: x[0]):
f.write('User ID %s top-10 results\n' % uid)
for iid, score in ratings:
f.write('Item ID %s\tscore %s\n' % (iid, str(score)))
f.write('\n')
# TODO - 3-4-5. Best Model
kfold = KFold(n_splits=5, random_state=0)
parameters_SVD = {
'n_factors': [50, 100, 200],
'n_epochs': [10, 50, 100, 200],
'biased': [True, False]
}
grid_SVD = GridSearchCV(surprise.SVD,
measures=['rmse'],
param_grid=parameters_SVD,
cv=kfold)
parameters_SVDpp = {
'n_factors': [50, 100, 200],
'n_epochs': [10, 50, 100, 200]
}
grid_SVDpp = GridSearchCV(surprise.SVDpp,
measures=['rmse'],
param_grid=parameters_SVDpp,
cv=kfold)
grid_SVD.fit(data)
grid_SVDpp.fit(data)
best_SVD_score = grid_SVD.best_score['rmse']
best_SVDpp_score = grid_SVDpp.best_score['rmse']
if best_SVD_score < best_SVDpp_score:
algo_name = 'SVD'
best_algo_mf = grid_SVD.best_estimator['rmse']
with_parameters = grid_SVD.best_params['rmse']
score = best_SVD_score
else:
algo_name = 'SVDpp'
best_algo_mf = grid_SVDpp.best_estimator['rmse']
with_parameters = grid_SVDpp.best_params['rmse']
score = best_SVDpp_score
print('The best MF algorithm is', algo_name, 'with', with_parameters,
'\nscore:', score)
print("Done.")
# defining the number of folds = 5
print("Performing splits...")
kf = sp.model_selection.KFold(n_splits=5, random_state=0)
print("Done.")
###
### PART 1.1
###
'''
application of all algorithms for recommendation made available by
“Surprise” libraries, according to their default configuration.
'''
algorithms = [sp.NormalPredictor(), sp.BaselineOnly(), sp.KNNBasic(),\
sp.KNNWithMeans(), sp.KNNWithZScore(), sp.KNNBaseline(),\
sp.SVD(), sp.SVDpp(), sp.NMF(), sp.SlopeOne(), sp.CoClustering()]
for elem in algorithms:
start_time = time.time()
algo = elem
sp.model_selection.cross_validate(algo, data, measures=['RMSE'], \
cv=kf, n_jobs = 2, verbose=True)
print("--- %s seconds ---" % (time.time() - start_time))
print()
###
### PART 1.2
###
'''
Improvement of the quality of both KNNBaseline and SVD methods,
by performing hyper-parameters tuning over 5-folds
import surprise as sp
from surprise import Dataset
from surprise.model_selection import cross_validate
import NetflixDataLoad
#for 100000 rows for fast processing
data = Dataset.load_from_df(
NetflixDataLoad.df_filterd[['Cust_Id', 'Movie_Id', 'Rating']][:100000])
n_folds = 5
for algo in [sp.SVD(), sp.SVDpp(), sp.KNNBasic(), sp.KNNWithMeans()]:
print(
cross_validate(algo,
data,
measures=['RMSE', 'MAE'],
cv=n_folds,
verbose=True))
# Output Example
# Evaluating RMSE, MAE of algorithm SVD on 5 split(s).
#
# Fold 1 Fold 2 Fold 3 Fold 4 Fold 5 Mean Std
# RMSE 0.9311 0.9370 0.9320 0.9317 0.9391 0.9342 0.0032
# MAE 0.7350 0.7375 0.7341 0.7342 0.7375 0.7357 0.0015
# Fit time 6.53 7.11 7.23 7.15 3.99 6.40 1.23
# Test time 0.26 0.26 0.25 0.15 0.13 0.21 0.06
}
mean_ap = []
precision = []
recall = []
fscore = []
normalized_DCG = []
mean_ap_train = []
precision_train = []
recall_train = []
fscore_train = []
normalized_DCG_train = []
for k_val in ks:
print(k_val)
algo = surprise.KNNWithMeans(k=k_val, sim_options=sim_options)
pr = 0
re = 0
fs = 0
ap = 0
nd = 0
pr_train = 0
re_train = 0
fs_train = 0
ap_train = 0
nd_train = 0
for trainset, testset in data.folds():
algo.train(trainset)
predictions_on_test = algo.test(testset)
precisions_test, recalls_test = precision_recall_at_k(
# A reader is needed but only the rating_scale param is requiered.
reader = Reader(rating_scale=(1, 5))
data = Dataset.load_from_df(train_df, reader)
sim_options = {'user_based': [False]}
results = []
# Iterate over all algorithms
for algorithm in [
SVD(),
surprise.NMF(),
surprise.SlopeOne(),
surprise.CoClustering(),
surprise.KNNBasic(sim_options=sim_options),
surprise.KNNWithMeans(sim_options=sim_options),
surprise.KNNWithZScore(sim_options=sim_options),
surprise.KNNBaseline(sim_options=sim_options),
surprise.NormalPredictor(),
surprise.BaselineOnly()
]:
# Get string of algname for naming a pickle file a useful name
alg_name = str(algorithm)
alg_name = alg_name[alg_name.find('.') + 1:]
alg_name = alg_name[alg_name.find('.') + 1:]
alg_name = alg_name[alg_name.find('.') + 1:]
alg_name = alg_name[:alg_name.find('object') - 1]
# Take a look at cross validation results to compare model types
print('\n\nModeling: {}\n'.format(str(alg_name)))
epochs=2,
validation_split=0.1,
shuffle=True)
y_pred = model.predict([df_hybrid_test['User'], df_hybrid_test['Movie'], test_tfidf])
y_true = df_hybrid_test['Rating'].values
rmse = np.sqrt(mean_squared_error(y_pred=y_pred, y_true=y_true))
print('\n\nTesting Result With Keras Hybrid Deep Learning: {:.4f} RMSE'.format(rmse))
# Load dataset into surprise specific data-structure
data = sp.Dataset.load_from_df(df_filterd[['User', 'Movie', 'Rating']].sample(20000), sp.Reader())
benchmark = []
# Iterate over all algorithms
for algorithm in [sp.SVD(), sp.SVDpp(), sp.SlopeOne(), sp.NMF(), sp.NormalPredictor(), sp.KNNBaseline(), sp.KNNBasic(), sp.KNNWithMeans(), sp.KNNWithZScore(), sp.BaselineOnly(), sp.CoClustering()]:
# Perform cross validation
results = cross_validate(algorithm, data, measures=['RMSE', 'MAE'], cv=3, verbose=False)
# Get results & append algorithm name
tmp = pd.DataFrame.from_dict(results).mean(axis=0)
tmp = tmp.append(pd.Series([str(algorithm).split(' ')[0].split('.')[-1]], index=['Algorithm']))
# Store data
benchmark.append(tmp)
# Store results
surprise_results = pd.DataFrame(benchmark).set_index('Algorithm').sort_values('test_rmse', ascending=False)
# Get data
data = surprise_results[['test_rmse', 'test_mae']]
for j in tp.columns[1:-1]:
tp[j] = tp[j].replace(np.nan, tpmean[j])
mb += [tp]
mb = pd.concat(mb, axis=0, ignore_index=True)
# Model Based
df = pd.read_csv(
'C:/Users/mayij/Desktop/DOC/GITHUB/MLGH/collaborative filtering/movielens.csv'
)
reader = surprise.Reader(rating_scale=(1, 5))
data = surprise.Dataset.load_from_df(df[['user', 'item', 'rating']], reader)
trainset = data.build_full_trainset()
# KNN
algo = surprise.KNNWithMeans(sim_options={
'name': 'cosine',
'user_based': False
})
algo.fit(trainset)
gs = surprise.model_selection.GridSearchCV(surprise.KNNWithMeans, {
'sim_options': {
'name': ['msd', 'cosine'],
'min_support': [3, 4, 5],
'user_based': [False, True]
}
},
measures=['rmse', 'mae'],
cv=3)
gs.fit(data)
print(gs.best_score['rmse'])
print(gs.best_params['rmse'])
