def run_on_soybean(file):
print("_______________________________")
print("Reading in Soybean data set...")
# Read in soybean data
df_soybean = pd.read_csv(file, header=None)
# Generate boolean classifiers (0 = not-Iris-virginica, 1 = Iris-virginica)
df_soybean = df_soybean.rename(columns={35: 'Class'})
df_soybean.loc[df_soybean["Class"] == "D1", "Class"] = 0
df_soybean.loc[df_soybean["Class"] == "D2", "Class"] = 0
df_soybean.loc[df_soybean["Class"] == "D3", "Class"] = 1
df_soybean.loc[df_soybean["Class"] == "D4", "Class"] = 1
# One hot encode breast data set for naive bayes
columns_to_encode = df_soybean.columns.values.tolist()
del columns_to_encode[35]
df_soybean_encoded = one_hot_encoder(df_soybean, columns_to_encode)
lg_averages = run_logistic_regression(df_soybean, 35)
nb_averages = run_naive_bayes(df_soybean_encoded, 35)
print("----------------------------")
print(f"Averages over 5 experiments")
print("----------------------------")
print(f"Logistic Regression Averages = {lg_averages}%")
print(f"Naive Bayes Averages = {nb_averages}%")
def run_on_glass(file, num_hidden):
"""
This function runs logistic regression and naive bayes classifier on the glass data set
:param file: input file
"""
print("_______________________________")
print("Reading in Glass data set...")
# Read in glass data
df_glass = pd.read_csv(file, header=None)
df_glass.columns = [
"Id", "RI", "Na", "Mg", "Al", "Si", "K", "Ca", "Ba", "Fe", "Class"
]
# This data set has no missing values, so we will skip that step
# Drop Id
df_glass = df_glass.drop('Id', axis=1)
# Encode the class
df_glass = one_hot_encoder(df_glass, ['Class'])
df_glass = df_glass.rename(columns={"Class_1": "Class"})
df_glass = df_glass.drop(
columns=['Class_2', 'Class_3', 'Class_5', 'Class_6', 'Class_7'])
averages = run_backpropagation(df_glass, 9, int(num_hidden))
print("----------------------------")
print(f"Averages over 5 experiments")
print("----------------------------")
print(f"Averages over all 5 experiments = {averages}%")
def run_on_glass(file):
"""
This function runs logistic regression and naive bayes classifier on the glass data set
:param file: input file
"""
print("_______________________________")
print("Reading in Glass data set...")
# Read in glass data
df_glass = pd.read_csv(file, header=None)
df_glass.columns = [
"Id", "RI", "Na", "Mg", "Al", "Si", "K", "Ca", "Ba", "Fe", "Class"
]
# This data set has no missing values, so we will skip that step
# Drop Id
df_glass = df_glass.drop('Id', axis=1)
# Encode the class
df_glass = one_hot_encoder(df_glass, ['Class'])
df_glass = df_glass.rename(columns={"Class_1": "Class"})
df_glass = df_glass.drop(
columns=['Class_2', 'Class_3', 'Class_5', 'Class_6', 'Class_7'])
print(df_glass.head())
# One hot code the data set for naive bayes
columns_to_encode = df_glass.columns.values.tolist()
del columns_to_encode[9]
df_glass_encoded = one_hot_encoder(df_glass, columns_to_encode)
df_glass_encoded = df_glass_encoded[
[c for c in df_glass_encoded if c not in ['Class']] + ['Class']]
lg_averages = run_logistic_regression(df_glass, 9)
nb_averages = run_naive_bayes(df_glass_encoded,
len(df_glass_encoded.columns) - 1)
print("----------------------------")
print(f"Averages over 5 experiments")
print("----------------------------")
print(f"Logistic Regression Averages = {lg_averages}%")
print(f"Naive Bayes Averages = {nb_averages}%")
def run_on_breast(file):
"""
This function runs logistic regression and naive bayes classifier on the breast data set, it encodes
the classes to Benign = 0, Malignant = 1, and removes missing values
:param file: input file
"""
print("_______________________________")
print("Reading in Breast data set...")
df_breast = pd.read_csv(file, header=None)
df_breast.columns = [
"Sample Id", "Clump Thickness", "Uniformity of Cell Size",
"Uniformity of Cell Shape", "Marginal Adhesion",
"Single Epithelial Cell Size", "Bare Nuclei", "Bland Chromatin",
"Normal Nucleoli", "Mitoses", "Class"
]
# Find missing values and remove them, since there are so few
# The documentation notes that there are16 missing values in group 1 and 6 denoted by '?'
# I found 16 values in Group 6
# Since there are so few missing values I dropped those rows
df_breast = df_breast[df_breast["Bare Nuclei"] != '?']
# Drop Sample Id
df_breast = df_breast.drop('Sample Id', axis=1)
# Generate boolean classifiers (0 = Benign, 1 = Malignant)
df_breast.loc[df_breast["Class"] == 2, "Class"] = 0
df_breast.loc[df_breast["Class"] == 4, "Class"] = 1
# One hot encode breast data set for naive bayes
columns_to_encode = df_breast.columns.values.tolist()
del columns_to_encode[9]
df_breast_encoded = one_hot_encoder(df_breast, columns_to_encode)
lg_averages = run_logistic_regression(df_breast, 9)
nb_averages = run_naive_bayes(df_breast_encoded, 9)
print("----------------------------")
print(f"Averages over 5 experiments")
print("----------------------------")
print(f"Logistic Regression Averages = {lg_averages}%")
print(f"Naive Bayes Averages = {nb_averages}%")
def run_on_iris(file):
"""
This function runs the SFS Algorithm wrapping Naive Bayes to reduce the feature set
and then runs a K-means clustering algorithm to cluster the data. To test k-means as
a classifier Naive Bayes is run again with the cluster labels generated by K-means.
:param file: The file name that includes the data
:return: Nothing
"""
print("_______________________________")
print("Reading in Iris data set...")
# Read in iris data
df_iris = pd.read_csv(file, header=None)
df_iris.columns = [
"sepal length in cm", "sepal width in cm", "petal length in cm",
"petal width in cm", "Class"
]
# This data set has no missing values, so we will skip that step
# One hot code the classes in order to use my previous naive bayes algorithm
df_iris_encoded = one_hot_encoder(df_iris, ["Class"])
# Split into test and training sets
x_test, x_train = split_test_train(df_iris_encoded)
x_test = x_test.reset_index(drop=True)
x_train = x_train.reset_index(drop=True)
# Run stepwise forward selection to reduce the feature set
print(
"Run Stepwise forward selection to reduce the feature set on Iris...")
print("All features...")
features = df_iris_encoded.columns.values.tolist()[0:4]
print(features)
sfs = StepwiseForwardSelection(features, x_train.iloc[:, 0:4],
x_test.iloc[:, 0:4],
x_train["Class_Iris-virginica"],
x_test["Class_Iris-virginica"], nb.learn,
nb.test)
optimized_feature_set = sfs.run()
cluster_and_classify(optimized_feature_set, x_test, x_train)
def run_on_glass(file):
"""
This function runs the SFS Algorithm wrapping Naive Bayes to reduce the feature set
and then runs a K-means clustering algorithm to cluster the data. To test k-means as
a classifier Naive Bayes is run again with the cluster labels generated by K-means.
:param file: The file name that includes the data
:return: Nothing
"""
print("_______________________________")
print("Reading in Glass data set...")
# Read in glass data
df_glass = pd.read_csv(file, header=None)
df_glass.columns = [
"Id", "RI", "Na", "Mg", "Al", "Si", "K", "Ca", "Ba", "Fe", "Class"
]
# This data set has no missing values, so we will skip that step
# Drop Id
df_glass_all = df_glass.drop('Id', axis=1)
# One hot code the classes in order to use my previous naive bayes algorithm
df_glass_encoded = one_hot_encoder(df_glass_all, ["Class"])
# Split into test and training sets
x_test, x_train = split_test_train(df_glass_encoded)
x_test = x_test.reset_index(drop=True)
x_train = x_train.reset_index(drop=True)
# Run stepwise forward selection to reduce the feature set
print("Running SFS on Glass data set...")
features = df_glass_all.columns.values.tolist()[0:9]
print("All features...")
print(features)
sfs = StepwiseForwardSelection(features, x_train.iloc[:, 0:9],
x_test.iloc[:, 0:9], x_train["Class_1"],
x_test["Class_1"], nb.learn, nb.test)
optimized_feature_set = sfs.run()
cluster_and_classify(optimized_feature_set, x_test, x_train)
def run_on_computer(file, k):
"""
This function runs k-nearest neighbors on the Computer Hardware data set
:param file: input file
"""
print("_______________________________")
print("Reading in Computer Hardware data set...")
df_computer = pd.read_csv(file, header=None)
print(df_computer.head())
df_computer.columns = [
"0", "1", "2", "3", "4", "5", "6", "7", "Class", "9"
]
# One hot encode categorical values
df_computer = one_hot_encoder(df_computer, ["0", "1"])
print(df_computer.head())
run_k_nearest_neighbor_experiments(df_computer,
k,
False,
classification=False)
def run_on_soybean(file, num_hidden):
print("_______________________________")
print("Reading in Soybean data set...")
# Read in soybean data
df_soybean = pd.read_csv(file, header=None)
# Generate boolean classifiers
df_soybean = df_soybean.rename(columns={35: 'Class'})
df_soybean.loc[df_soybean["Class"] == "D1", "Class"] = 0
df_soybean.loc[df_soybean["Class"] == "D2", "Class"] = 0
df_soybean.loc[df_soybean["Class"] == "D3", "Class"] = 1
df_soybean.loc[df_soybean["Class"] == "D4", "Class"] = 1
# One hot encode breast data set for naive bayes
columns_to_encode = df_soybean.columns.values.tolist()
del columns_to_encode[35]
df_soybean_encoded = one_hot_encoder(df_soybean, columns_to_encode)
averages = run_backpropagation(df_soybean, 35, int(num_hidden))
print("----------------------------")
print(f"Averages over 5 experiments")
print("----------------------------")
print(f"Averages over all 5 experiments = {averages}%")