def recommand(self):
sim_options = {
'name': 'pearson_baseline',
'shrinkage': 0 # no shrinkage
}
best_model = knns.KNNWithMeans(k=20, sim_options=sim_options)
t_values, precisions_knn, recall_knn = self.test_with_t_and_k(
best_model, msg='KNN')
best_model = matrix_factorization.NMF(n_factors=20, biased=False)
t_values, precisions_nmf, recall_nmf = self.test_with_t_and_k(
best_model, msg='NMF')
best_model = SVD(20)
t_values, precisions_svd, recall_svd = self.test_with_t_and_k(
best_model, msg='SVD')
plt.plot(t_values, precisions_knn, label='precisions_knn')
plt.plot(t_values, precisions_nmf, label='precisions_nmf')
plt.plot(t_values, precisions_svd, label='precisions_svd')
plt.plot(t_values, recall_knn, label='recall_knn')
plt.plot(t_values, recall_nmf, label='recall_nmf')
plt.plot(t_values, recall_svd, label='recall_svd')
plt.xlabel('t_value')
plt.ylabel('percent')
plt.legend(loc="best")
plt.show()
def run_and_test_all_models(self):
step_size = 2
# KNN
sim_options = {
'name': 'pearson_baseline',
'shrinkage': 0 # no shrinkage
}
algo = knns.KNNWithMeans
args = {'sim_options': sim_options}
best_model = knns.KNNWithMeans(k=20, sim_options=sim_options)
roc_auc_KNN = self.run_and_test_model(algo, args, best_model,
(2, 101, step_size), 'KNN')
# # NMF
algo = matrix_factorization.NMF
args = {'biased': False}
best_model = matrix_factorization.NMF(n_factors=20, biased=False)
roc_auc_NMF = self.run_and_test_model(algo, args, best_model,
(2, 51, step_size), 'NMF')
# SVD
algo = matrix_factorization.SVD
args = {}
best_model = SVD(20)
roc_auc_SVD = self.run_and_test_model(algo, args, best_model,
(2, 51, step_size), 'SVD')
# all
for i in range(len(roc_auc_KNN)):
plt.plot(roc_auc_KNN[i][0],
roc_auc_KNN[i][1],
color='blue',
linewidth=2.0,
label='KNN')
plt.plot(roc_auc_NMF[i][0],
roc_auc_NMF[i][1],
color='blue',
linewidth=2.0,
label='NMF')
plt.plot(roc_auc_SVD[i][0],
roc_auc_SVD[i][1],
color='blue',
linewidth=2.0,
label='SVD')
plt.plot([0, 1], [0, 1], color='yellow', linewidth=2.0)
plt.xlabel('FPR')
plt.ylabel('TPR')
plt.legend(loc="lower right")
plt.show()
# NaiveFilter
self.run_naive_filter(msg='normal')
self.run_naive_filter(test_filter=trimPopular, msg='trimPopular')
self.run_naive_filter(test_filter=trimUnpopular, msg='trimUnpopular')
self.run_naive_filter(test_filter=trimHighVariance,
msg='trimHighVariance')
def non_negative_matrix_factorization(self):
algo = matrix_factorization.NMF(20)
model = algo.fit(self.data.build_full_trainset())
U = model.pu
V = model.qi
print U.shape
print V.shape
top_ten = V[:, 1].argsort()[::-1][:10]
movies = {}
genre = []
f = open('recommand/ml-latest-small/movies.csv')
movies_reader = csv.reader(f)
for movie_entry in movies_reader:
movies[movie_entry[0]] = movie_entry[2]
f.close()
for i in top_ten:
for j in movies:
if j != "movieId" and int(i) == int(j):
genre.append([i, movies[j]])
print(genre)
def _init_surprise_model(self):
return matrix_factorization.NMF(n_factors=self._n_factors, random_state=self._random_state,
n_epochs=self._n_epochs, reg_pu=self._reg_pu, reg_qi=self._reg_qi)