def modelPrompt(x_train, y_train, x_standard, y, classifier, name):
while True:
value = input(
"Please choose the maximum number of features to select: ")
try:
k_value = int(value)
except ValueError:
print("The input you entered is not an integer. Please try again.")
else:
if k_value > 0:
result = feature_selection.modelFeatSelect(
x_train, y_train, classifier, k_value)
print(result)
if not result.empty:
features = list(result['Feature'])
title = "Learning Curves (" + name + ", log regression model)"
x_selected = x_standard[features]
x_train_selected = x_train[features]
logreg = LogisticRegression(C=1)
logreg.fit(x_train_selected, y_train)
evaluate_model.plot_learning_curve(logreg,
title,
x_selected,
y,
cv=5)
break
else:
print(
"The integer you entered is not positive. Please try again."
)
def uniPrompt(x_train, y_train, x_standard, y):
print("Starting Univariance Feature Selection: ")
while True:
value = input("Please choose the number of features to select: ")
try:
k_value = int(value)
except ValueError:
print("The input you entered is not an integer. Please try again.")
else:
if k_value > 0:
result = feature_selection.univariance(x_train, y_train,
k_value)
print(result)
features = list(result['Feature'])
title = "Learning Curves (univariance, log regression model)"
x_selected = x_standard[features]
x_train_selected = x_train[features]
logreg = LogisticRegression(C=1)
logreg.fit(x_train_selected, y_train)
evaluate_model.plot_learning_curve(logreg,
title,
x_selected,
y,
cv=5)
break
else:
print(
"The integer you entered is not positive. Please try again."
)
def evalModelPrompt(x_train, x_test, y_train, y_test, x_standard, y,
classifier, name):
print("Find the best number of features using " + name + "... ")
selected_features = feature_selection.getBestK(x_train, x_test, y_train,
y_test, classifier, name)
title = "Learning Curves with " + name + " (use selected features)"
x_selected = x_standard[list(selected_features)]
x_train_selected = x_train[list(selected_features)]
classifier.fit(x_train_selected, y_train)
evaluate_model.plot_learning_curve(classifier, title, x_selected, y, cv=5)
def RFEPrompt(x_train, y_train, x_standard, y, classifier, name):
print("Starting RFE: ")
result = feature_selection.RFEFeatSelect(x_train, y_train, classifier)
print(result)
if not result.empty:
features = list(result['Feature'])
title = "Learning Curves with " + name + " (use current feature subset)"
x_selected = x_standard[features]
x_train_selected = x_train[features]
classifier.fit(x_train_selected, y_train)
evaluate_model.plot_learning_curve(classifier,
title,
x_selected,
y,
cv=5)
def main():
print("Begin Operation... ")
print(
"Notice: Non-numeric data besides the target class (can be categorical) will be ignored... "
)
# get path to csv file
csv_path = None
while True:
csv_path = input("Please provide a path to a csv file: ")
if os.path.exists(csv_path) == False:
print("The path that you provided is incorrect. Please try again.")
elif os.path.isfile(csv_path) == False:
print(
"The path that you provided is not a file. Please try again.")
elif csv_path.endswith('.csv') == False:
print(
"The path that you provided is not a csv file. Please try again."
)
else:
break
df = pd.read_csv(csv_path, index_col=0)
# get target class
while True:
target_name = input("Please provide the name of the target class: ")
if target_name in df.head():
x, y = preprocessing.separateVars(df, target_name)
break
else:
print(
"The name that you provided doesn't exist in this csv file. Please try again."
)
# get target class encoding
y = encodePrompt(y, target_name)
# remove non numeric data from feature columns
x = preprocessing.cleanData(x)
print("Non-numeric columns and columns with missing values removed.")
# normalize data
x_standard = preprocessing.standardization(x)
print("Data Normalized (z-score).")
# visualize data
visualPrompt(x_standard, y)
# prompt for feature transformation
x_standard, y = transformationPrompt(x_standard, y)
x_standard = preprocessing.standardization(x_standard)
answer = input("Would you like to perform feature selection? (y/n) ")
if answer == 'y':
# split dataset to train and test
x_train, x_test, y_train, y_test = train_test_split(x_standard,
y,
test_size=0.3,
random_state=0)
method = input(
"Would you like to perform classification or regression? (c/r) ")
if method == 'c':
# find optimal classifier
print("Begin to test classifiers... ")
classifier1, name1, classifier2, name2 = feature_selection.optimalClassifier(
x_train, x_test, y_train, y_test)
# plot learning curve of current classifier
title = "Learning Curves (" + name1 + ")"
evaluate_model.plot_learning_curve(classifier1,
title,
x_standard,
y,
cv=5)
title = "Learning Curves (" + name2 + ")"
evaluate_model.plot_learning_curve(classifier2,
title,
x_standard,
y,
cv=5)
answer = input(
"Do you want to find the best number of feature using " +
name1 + "? (y/n) ")
if answer == 'y':
evalModelPrompt(x_train, x_test, y_train, y_test, x_standard,
y, classifier1, name1)
else:
print("Skip for " + name1 + ".")
answer = input(
"Do you want to find the best number of feature using " +
name2 + "? (y/n) ")
if answer == 'y':
evalModelPrompt(x_train, x_test, y_train, y_test, x_standard,
y, classifier2, name2)
else:
print("Skip for " + name2 + ".")
elif method == 'r':
print("Begin to test regression models... ")
model1, name1, model2, name2 = feature_selection.optimalRegression(
x_train, x_test, y_train, y_test)
# plot learning curve of current classifier
title = "Learning Curves (" + name1 + ")"
evaluate_model.plot_learning_curve(model1,
title,
x_standard,
y,
cv=5)
title = "Learning Curves (" + name2 + ")"
evaluate_model.plot_learning_curve(model2,
title,
x_standard,
y,
cv=5)
else:
print("Your input doesn't match classification or regression.")
# feature selection
while True:
method = input(
"Please choose a feature selection method(RFE/univariance/lasso/optimal_classifier/MRMR): "
)
if method == 'RFE':
print("Using " + name1)
RFEPrompt(x_train, y_train, x_standard, y, classifier1, name1)
print("Using " + name2)
RFEPrompt(x_train, y_train, x_standard, y, classifier2, name2)
break
elif method == 'univariance':
uniPrompt(x_train, y_train, x_standard, y)
break
elif method == 'lasso':
print("Using lasso")
modelPrompt(x_train, y_train, x_standard, y, LassoCV(cv=5),
'Lasso')
break
elif method == 'optimal_classifier':
print("Using " + name1)
modelPrompt(x_train, y_train, x_standard, y, classifier1,
name1)
print("Using " + name2)
modelPrompt(x_train, y_train, x_standard, y, classifier2,
name2)
break
elif method == 'MRMR':
result = feature_selection.mrmr(x_standard, y)
print(result)
break
else:
print(
"The method you provide is not an option. Please try again. "
)
else:
print("Feature Selection is skipped. ")
print("End of Operation.")