def baseline_global_mean(train, test, training=False):
mean = train.Rating.mean()
pred_test = test.copy()
pred_test.Rating = mean
if training:
rmse = compute_rmse(pred_test, test)
print("baseline_global_mean: {}".format(rmse))
return pred_test
def movie_mean_user_standardize(train, test, training=False):
stand_train = user_standardize(train)
stand_pred_test = baseline_movie_mean(stand_train, test)
#recover from the standardized predicted test rating
pred_test = user_standardize_recover(train, stand_pred_test)
#compute the rmse
if training:
rmse = compute_rmse(pred_test, test)
print("movie_mean_user_std: {}".format(rmse))
return pred_test
def baseline_movie_median(train, test, training=False):
median_per_movie = train.groupby('Movie').median().Rating
pred_test = test.copy()
def predict(sub_df):
sub_df['Rating'] = median_per_movie[sub_df.iloc[0, 1]]
return sub_df
pred_test = pred_test.groupby('Movie').apply(predict)
if training:
rmse = compute_rmse(pred_test, test)
print("baseline_movie_mean: {}".format(rmse))
return pred_test
def movie_median_user_habit_standardize(train, test, training=False):
#standardize the rating according to per user mean and variance
stand_train = user_habit_standardize(train)
#predict the standardized test rating
stand_pred_test = baseline_movie_median(stand_train, test)
#recover from the standardized predicted test rating
pred_test = user_habit_standardize_recover(train, stand_pred_test)
#compute the rmse
if training:
rmse = compute_rmse(pred_test, test)
print("movie_median_user_habit_std: {}".format(rmse))
return pred_test
def movie_mean_user_habit_standardize(train, test, training=False):
#standardize the rating according to per user habit
pred_test = test.copy()
pred_test.Rating = pred_test.Rating.apply(lambda x: float(x))
stand_train = user_habit_standardize(train)
#predict the standardized test rating
stand_pred_test = baseline_movie_mean(stand_train, test)
#recover from the standardized predicted test rating
pred_test = user_habit_standardize_recover(train, stand_pred_test)
#compute the rmse
if training:
rmse = compute_rmse(pred_test, test)
print("movie_mean_user_habit_std: {}".format(rmse))
return pred_test
def movie_median_user_habit(train, test, training=False):
habit = user_habit(train)
median_per_movie = train.groupby('Movie').median().Rating
pred_test = test.copy()
pred_test['Rating'] = pred_test['Rating'].apply(lambda x: float(x))
def predict(x):
x['Rating'] = median_per_movie[x['Movie']] + habit[x['User']]
return x
pred_test = pred_test.apply(predict, axis=1)
pred_test['User'] = pred_test['User'].apply(lambda x: int(x))
pred_test['Movie'] = pred_test['Movie'].apply(lambda x: int(x))
if training:
rmse = compute_rmse(pred_test, test)
print("movie_median_user_habit: {}".format(rmse))
return pred_test