def save_oos_pred(self) :
"""
For all the models specified in self.d_model,
get OOS predictions to get LV1 train features
Save them to a csv file
"""
#Loop over estimators
for model_name in self.d_model.keys() :
#Only train if desired
if self.d_model[model_name]["train"]:
# Get data
feat_choice = self.d_model[model_name]["feat_choice"]
X, y, Id = prep.prepare_lv1_data(feat_choice, "train")
y_keep = y.copy() # keep a copy of y to deal with classifier which may modify y
print "Compute OOS pred for model: ", model_name
# Compute OOS preds
y_pred_oos, list_col = self._get_oos_preds(X, y, Id, model_name)
#Save predictions to csv file with pandas
Id = np.reshape(Id, (Id.shape[0],1))
y_keep = np.reshape(y_keep, (y_keep.shape[0],1))
y_pred_oos = np.reshape(y_pred_oos, (y_pred_oos.shape[0], len(list_col)-2))
data = np.hstack((Id, y_pred_oos, y_keep))
df = pd.DataFrame(data, columns=list_col)
df[["Id", "Response"]] = df[["Id", "Response"]].astype(int)
df = df.sort_values("Id")
# Specify output file name
out_name = "./Data/Level1_model_files/Train/%s_%s_train.csv" % (model_name, feat_choice)
# Special case for KNN : specify n_neighbors in out_name
if "knn" in model_name :
out_name = "./Data/Level1_model_files/Train/%s_%s_%s_train.csv" % \
(model_name, feat_choice, self.d_model[model_name]["n_neighbors"])
df.to_csv(out_name, index = False, float_format='%.3f')
def save_test_pred(self):
"""
For all the models specified in self.d_model,
get LV1 test features
Save them to a csv file
"""
if not os.path.exists("./Data/Level1_model_files/Test/"):
os.makedirs("./Data/Level1_model_files/Test/")
#Loop over estimators
for model_name in self.d_model.keys():
#Only train if desired
if self.d_model[model_name]["train"]:
# Get data
feat_choice = self.d_model[model_name]["feat_choice"]
X_train, y_train, Id_train = prep.prepare_lv1_data(
feat_choice, "train")
X_test, dummy, Id_test = prep.prepare_lv1_data(
feat_choice, "test")
print "Compute test pred for model: ", model_name
# Compute test preds
y_pred_test, list_col = self._get_test_preds(
X_train, y_train, Id_train, X_test, Id_test, model_name)
#Save predictions to csv file with pandas
Id_test = np.reshape(Id_test, (Id_test.shape[0], 1))
y_pred_test = np.reshape(
y_pred_test, (y_pred_test.shape[0], len(list_col) - 1))
data = np.hstack((Id_test, y_pred_test))
df = pd.DataFrame(data, columns=list_col)
df["Id"] = df["Id"].astype(int)
df = df.sort_values("Id")
# Specify output file name
out_name = "./Data/Level1_model_files/Test/%s_%s_test.csv" % (
model_name, feat_choice)
# Special case for KNN : specify n_neighbors in out_name
if "knn" in model_name:
out_name = "./Data/Level1_model_files/Test/%s_%s_%s_test.csv" % \
(model_name, feat_choice, self.d_model[model_name]["n_neighbors"])
df.to_csv(out_name, index=False, float_format='%.3f')
def save_test_pred(self) :
"""
For all the models specified in self.d_model,
get LV1 test features
Save them to a csv file
"""
if not os.path.exists("./Data/Level1_model_files/Test/"):
os.makedirs("./Data/Level1_model_files/Test/")
#Loop over estimators
for model_name in self.d_model.keys() :
#Only train if desired
if self.d_model[model_name]["train"]:
# Get data
feat_choice = self.d_model[model_name]["feat_choice"]
X_train, y_train, Id_train = prep.prepare_lv1_data(feat_choice, "train")
X_test,dummy, Id_test = prep.prepare_lv1_data(feat_choice, "test")
print "Compute test pred for model: ", model_name
# Compute test preds
y_pred_test, list_col = self._get_test_preds(X_train, y_train, Id_train, X_test, Id_test, model_name)
#Save predictions to csv file with pandas
Id_test = np.reshape(Id_test, (Id_test.shape[0],1))
y_pred_test = np.reshape(y_pred_test, (y_pred_test.shape[0], len(list_col)-1))
data = np.hstack((Id_test, y_pred_test))
df = pd.DataFrame(data, columns=list_col)
df["Id"] = df["Id"].astype(int)
df = df.sort_values("Id")
# Specify output file name
out_name = "./Data/Level1_model_files/Test/%s_%s_test.csv" % (model_name, feat_choice)
# Special case for KNN : specify n_neighbors in out_name
if "knn" in model_name :
out_name = "./Data/Level1_model_files/Test/%s_%s_%s_test.csv" % \
(model_name, feat_choice, self.d_model[model_name]["n_neighbors"])
df.to_csv(out_name, index = False, float_format='%.3f')
def save_oos_pred(self):
"""
For all the models specified in self.d_model,
get OOS predictions to get LV1 train features
Save them to a csv file
"""
#Loop over estimators
for model_name in self.d_model.keys():
#Only train if desired
if self.d_model[model_name]["train"]:
# Get data
feat_choice = self.d_model[model_name]["feat_choice"]
X, y, Id = prep.prepare_lv1_data(feat_choice, "train")
y_keep = y.copy(
) # keep a copy of y to deal with classifier which may modify y
print "Compute OOS pred for model: ", model_name
# Compute OOS preds
y_pred_oos, list_col = self._get_oos_preds(
X, y, Id, model_name)
#Save predictions to csv file with pandas
Id = np.reshape(Id, (Id.shape[0], 1))
y_keep = np.reshape(y_keep, (y_keep.shape[0], 1))
y_pred_oos = np.reshape(
y_pred_oos, (y_pred_oos.shape[0], len(list_col) - 2))
data = np.hstack((Id, y_pred_oos, y_keep))
df = pd.DataFrame(data, columns=list_col)
df[["Id", "Response"]] = df[["Id", "Response"]].astype(int)
df = df.sort_values("Id")
# Specify output file name
out_name = "./Data/Level1_model_files/Train/%s_%s_train.csv" % (
model_name, feat_choice)
# Special case for KNN : specify n_neighbors in out_name
if "knn" in model_name:
out_name = "./Data/Level1_model_files/Train/%s_%s_%s_train.csv" % \
(model_name, feat_choice, self.d_model[model_name]["n_neighbors"])
df.to_csv(out_name, index=False, float_format='%.3f')
