def matrix_factorization_param(data_cv):
# Iterate over all algorithms
benchmark = []
for algorithm in [
SVD(),
SVDpp(),
NMF(),
SlopeOne(),
NormalPredictor(),
CoClustering()
]:
# Perform cross validation
results = model_selection.cross_validate(algorithm,
data_cv,
measures=['RMSE', 'MAE'],
cv=5,
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)
rmse = pd.DataFrame(benchmark).set_index('Algorithm').sort_values(
'test_mae')
#print(rmse)
# Parameter grid
param_grid = {
'n_factors': [100, 150, 200],
'n_epochs': [20, 40],
'lr_all': [0.001, 0.005, 0.008],
'reg_all': [0.075, 0.1, 0.15]
}
algorithm_gs = model_selection.GridSearchCV(SVD,
param_grid,
measures=['rmse'],
cv=5,
n_jobs=-1)
algorithm_gs.fit(data_cv)
# best parameters for a model with the lowest rmse
best_algo = algorithm_gs.best_estimator['rmse']
return best_algo
rating_scale=(1, 5),
skip_lines=1)
data_train = Dataset.load_from_file(file_path, reader=reader)
# %% Hyper parameter tuning and CV analysis
# Algorithm: SVD
Hyper_Params = {
'n_epochs': [10],
'n_factors': [50, 100, 150, 200],
'biased': [False],
'lr_all': [0.005],
'reg_all': [0.01, 0.1, 0.3, 1.0]
}
Train_CV = Grid_Search_Result = model_selection.GridSearchCV(
SVD, Hyper_Params, measures=['rmse', 'mae'], cv=3, n_jobs=3)
Train_CV.fit(data_train)
# %% Figures
plt.figure(figsize=(20, 12))
plt.rcParams.update({'font.size': 12})
plt.plot(Train_CV.cv_results['param_reg_all'],
Train_CV.cv_results['mean_test_rmse'], '.k')
plt.xscale('log')
plt.xlabel('Regularization Parameter ($\lambda$)')
plt.ylabel('RMSE')
plt.grid()
plt.title('3-Fold CV - Regularization Parameter ($\lambda$)')
plt.savefig('3_fold_CV_Reg_Param.png')
reader = Reader(line_format='item user rating', sep=',',
rating_scale=(1, 5), skip_lines=1)
data_train = Dataset.load_from_file(file_path, reader=reader)
# %% Hyper parameter tuning and CV analysis
Hyper_Params = {'bsl_options':
{'method': ['als'],
'n_epochs': [20],
'reg_u': [1, 3, 10],
'reg_i': [1, 3, 10]}}
start = time.time()
Train_CV = Grid_Search_Result = model_selection.GridSearchCV(BaselineOnly,
Hyper_Params,
measures=['rmse'],
cv=3, n_jobs=3,
return_train_measures=True,
joblib_verbose=3)
Train_CV.fit(data_train)
end = time.time()
print("***********************************************")
print("Exe time:")
print(end - start)
# %% Figures
reg_i = []
reg_u = []
for i in Train_CV.cv_results['param_bsl_options']:
user_ratings.sort(key=lambda x: x[1], reverse=True)
top_n[uid] = user_ratings[:n]
return top_n
movie_train = Dataset.load_builtin('ml-100k')
print(movie_train.raw_ratings)
svd_estimator = surprise.SVD
svd_grid = {
'n_factors': [50, 100, 150],
'reg_all': [0, 0.4, 0.6],
'biased': ['True', 'False']
}
gs = model_selection.GridSearchCV(svd_estimator,
svd_grid,
measures=['rmse'],
cv=3)
gs.fit(movie_train)
print(gs.best_score['rmse'])
print(gs.best_params['rmse'])
final_estimator = gs.best_estimator['rmse']
#build final model on entire train data
movie_train = movie_train.build_full_trainset()
final_estimator.fit(movie_train)
movie_test = movie_train.build_anti_testset()
predictions = final_estimator.test(movie_test)
top_n = get_top_n(predictions, n=3)
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'])
print(knn_grid_estimator.best_params['rmse'])
results_df = pd.DataFrame.from_dict(knn_grid_estimator.cv_results)
final_model = knn_grid_estimator.best_estimator['rmse']
#build final model using best parameters from grid search
trainSet = movie_train.build_full_trainset()
final_model.fit(trainSet)
rows = csv.reader(open('F:/test_v2.csv'))
rows = list(rows)
rows.pop(0)
return data
movie_train = read_train_data(
'C:\\Users\\Algorithmica\\Downloads\\train_v2.csv')
print(movie_train.raw_ratings)
knn_grid = {
'k': [10, 20],
'sim_options': {
'name': ['cosine'],
'min_support': [1, 5],
'user_based': [False]
}
}
gs = model_selection.GridSearchCV(KNNBasic, knn_grid, measures=['rmse'], cv=3)
gs.fit(movie_train)
print(gs.best_score['rmse'])
print(gs.best_params['rmse'])
results_df = pd.DataFrame.from_dict(gs.cv_results)
algo = gs.best_estimator['rmse']
trainSet = movie_train.build_full_trainset()
algo.fit(trainSet)
rows = csv.reader(open('F:/test_v2.csv'))
rows = list(rows)
rows.pop(0)
f = open('F:/submission.csv', 'w', newline='')
writer = csv.writer(f)
writer.writerow(['ID', 'rating'])
