def base_dt(ver, inname, outname, testname):
X, Y = data_manip.read_indexed(inname)
X_test, Y_test = data_manip.read_indexed(testname)
dtc = tree.DecisionTreeClassifier()
dtc.fit(X, Y)
output_arr = [dtc.predict([X])[0] for X in X_test]
data_manip.write_indexed(outname, pd.DataFrame({'index': output_arr}))
calculate_metrics(Y_test, output_arr, ver, outname)
def gnb_predictor(ver, inname, outname, testname):
X, Y = data_manip.read_indexed(inname)
X_test, Y_test = data_manip.read_indexed(testname)
gnb = GaussianNB()
gnb.fit(X, Y)
output_arr = [gnb.predict([X])[0] for X in X_test]
data_manip.write_indexed(outname, pd.DataFrame({'index': output_arr}))
calculate_metrics(Y_test, output_arr, ver, outname)
def default_perceptron(ver, inname, outname, testname):
X, Y = data_manip.read_indexed(inname)
X_test, Y_test = data_manip.read_indexed(testname)
per = Perceptron() # Default params
per.fit(X, Y)
output_arr = [per.predict([X])[0] for X in X_test]
data_manip.write_indexed(outname, pd.DataFrame({'index': output_arr}))
calculate_metrics(Y_test, output_arr, ver, outname)
def base_multi_layered_perceptron(ver, inname, outname, testname):
X, Y = data_manip.read_indexed(inname)
X_test, Y_test = data_manip.read_indexed(testname)
mlp = MLPClassifier(hidden_layer_sizes=(100, ),
activation="logistic",
solver='sgd',
max_iter=400)
mlp.fit(X, Y)
output_arr = [mlp.predict([X])[0] for X in X_test]
data_manip.write_indexed(outname, pd.DataFrame({'index': output_arr}))
calculate_metrics(Y_test, output_arr, ver, outname)
def best_dt(ver, inname, outname, testname):
"""
splitting criterion: entropy
maximum depth of the tree: 85
minimum number of samples to split an internal node: experiment with values of your choice
minimum impurity decrease: experiment with values of your choice
class weight: None and balanced
"""
X, Y = data_manip.read_indexed(inname)
X_test, Y_test = data_manip.read_indexed(testname)
dtc = tree.DecisionTreeClassifier()
dtc.max_depth = None
dtc.criterion = "entropy"
dtc.min_samples_split = 5
dtc.min_impurity_decrease = 0.00025
dtc.class_weight = None
dtc.fit(X, Y)
output_arr = [dtc.predict([X])[0] for X in X_test]
data_manip.write_indexed(outname, pd.DataFrame({'index': output_arr}))
calculate_metrics(Y_test, output_arr, ver, outname)
def best_multi_layered_perceptron(ver, inname, outname, testname):
X, Y = data_manip.read_indexed(inname)
X_test, Y_test = data_manip.read_indexed(testname)
# for both datasets this is the opmimal hyper parameters
mlp = MLPClassifier(hidden_layer_sizes=(50, 50),
activation="relu",
solver='adam',
max_iter=400)
# THIS IS USED TO FIND THE GOOD HYPERPARAMS
# mlp = MLPClassifier(max_iter=400)
# parameters = {
# 'hidden_layer_sizes': [(30,50), (50,50)],
# 'activation': ['identity', 'logistic', 'tanh', 'relu'],
# 'solver': ['sgd','adam']
# }
# gridMLP = GridSearchCV(mlp, parameters)
# gridMLP.fit(X, Y)
# print("BEST PARAMS FOR: "+str(ver))
# print(gridMLP.get_params())
# output_arr = [gridMLP.predict([X])[0] for X in X_test]
mlp.fit(X, Y)
output_arr = [mlp.predict([X])[0] for X in X_test]
data_manip.write_indexed(outname, pd.DataFrame({'index': output_arr}))
calculate_metrics(Y_test, output_arr, ver, outname)
def calculate_distribution(ver):
X, Y = data_manip.read_indexed("./data/train_" + str(ver) + ".csv")
df = pd.DataFrame({'index': Y})
return df['index'].value_counts()