def predict_base_mlp(input_data, validation_data, test_data, filename):
inputX = input_data.iloc[:, 0:1024] # Binary features (first 1024 columns)
inputY = input_data.iloc[:,
1024] # Index representing the class (last column)
# Create a decision tree fit using the training data
basemlp = MLPClassifier(100, activation='logistic', solver='sgd')
basemlp.fit(inputX, inputY)
# Validate decision tree. But we aren't calibrating any parameters so this is optional.
validX = validation_data.iloc[:, 0:1024]
validY = validation_data.iloc[:, 1024]
score = basemlp.score(validX, validY)
print(f'Score on validation set: {score}')
# Predict using the test data
testX = test_data.iloc[:, 0:1024]
testY = test_data.iloc[:, 1024]
predictions = basemlp.predict(testX)
score2 = basemlp.score(testX, testY)
# Print the predictions to CSV
output_df = pd.DataFrame(predictions)
output_df.transpose()
fileutils.write_output(filename, output_df)
return predictions, testY
def predict_gnb(input_data, validation_data, test_data, filename):
inputX = input_data.iloc[:, 0:1024] # Binary features (first 1024 columns)
inputY = input_data.iloc[:,
1024] # Index representing the class (last column)
# Create a GNB fit using the training data
gnb = GaussianNB()
gnb.fit(inputX, inputY)
# Validate GNB. But we aren't calibrating any parameters so this is optional.
validX = validation_data.iloc[:, 0:1024]
validY = validation_data.iloc[:, 1024]
score = gnb.score(validX, validY)
print(f'Score: {score}')
# Predict using the test data
testX = test_data.iloc[:, 0:1024]
testY = test_data.iloc[:, 1024]
predictions = gnb.predict(testX)
# Print the predictions to CSV
output_df = pd.DataFrame(predictions)
output_df.transpose()
fileutils.write_output(filename, output_df)
return predictions, testY
def predict_best_mlp(input_data, validation_data, test_data, filename):
inputX = input_data.iloc[:,0:1024] # Binary features (first 1024 columns)
inputY = input_data.iloc[:,1024] # Index representing the class (last column)
# Create a mlp fit using the training data
basemlp = MLPClassifier(**best_mlp_params(input_data))
basemlp.fit(inputX, inputY)
# Validate the mlp.
validX = validation_data.iloc[:, 0:1024]
validY = validation_data.iloc[:,1024]
score = basemlp.score(validX, validY)
print(f'Score on validation set: {score}')
# Predict using the test data
testX = test_data.iloc[:,0:1024]
testY = test_data.iloc[:,1024]
predictions = basemlp.predict(testX)
score2 = basemlp.score(testX, testY)
# Print the predictions to CSV
output_df = pd.DataFrame(predictions)
output_df.transpose()
fileutils.write_output(filename, output_df)
return predictions, testY
def predict_perceptron(input_data, validation_data, test_data, filename):
inputX = input_data.iloc[:, 0:1024] # Binary features (first 1024 columns)
inputY = input_data.iloc[:,
1024] # Index representing the class (last column)
# Create a perceptron fit using the training data
perceptron = Perceptron()
perceptron.fit(inputX, inputY)
# Validate perceptron
validX = validation_data.iloc[:, 0:1024]
validY = validation_data.iloc[:, 1024]
score = perceptron.score(validX, validY)
print(f'Score on validation set: {score}')
# Predict using the test data
testX = test_data.iloc[:, 0:1024]
testY = test_data.iloc[:, 1024]
predictions = perceptron.predict(testX)
score2 = perceptron.score(testX, testY)
# Print the predictions to CSV
output_df = pd.DataFrame(predictions)
output_df.transpose()
fileutils.write_output(filename, output_df)
return predictions, testY