Boosting tree - Xgboost + GridSearchCV

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

from sklearn.model_selection import train_test_split

from sklearn.metrics import r2_score, mean_squared_error

from xgboost import XGBRegressor

from sklearn.model_selection import GridSearchCV

# Install ...

!pip install xgboost

# Data load

df = pd.read_csv('./data/xxx.csv')

# Target data separation

y = df['y']

X = df.drop('y', axis = 1)

# Split train : test data set = 7:3 and set random seed

X_train, X_test, y_train, y_test = train_test_split(X,

                                                    y,

                                                    test_size = 0.3,

                                                    random_state = 42)

# Define a model structure

xgbr = XGBRegressor(random_state = 42)

# Set the range of parameters

params = {

    'max_depth' : [3, 5, 7, 9],

    'n_estimators' : [50, 70, 90, 100],

    'learning_rate' : [0.03, 0.01],

    'subsample' : [0.7, 0.8, 0.9],

    'colsample_bytree' : [0.7, 0.8, 0.9]

    # 'reg_alpha' : [0, 1],

    # 'reg_lambda' : [0, 1]

}

# Set the condition of grid search model

grid = GridSearchCV(estimator = xgbr,

                   param_grid = params,

                   scoring = 'r2',

                   n_jobs = -1, # 최대 사용

                   verbose = 2,

                   cv = 5)

# Train the grid search model

grid.fit(X_train, y_train)

# Confirm the optimal parameters

grid.best_params_

# Model prediction using those optimal parameters

best_pred = grid.predict(X_test)

# Error metric

print("r2 score:", r2_score(y_test, best_pred))

print("RMSE:", np.sqrt(mean_squared_error(y_test, best_pred)))

Decision tree regressor

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

from sklearn.model_selection import train_test_split

from sklearn.tree import DecisionTreeRegressor

from sklearn.metrics import r2_score, mean_squared_error

# Data load

df = pd.read_csv('./data/xxx.csv')

# Target data separation

y = df['y']

X = df.drop('y', axis = 1)

# Split train : test data set = 7:3 and set random seed

X_train, X_test, y_train, y_test = train_test_split(X,

                                                    y,

                                                    test_size = 0.3,

                                                    random_state = 42)

# Define a model structure

dtr = DecisionTreeRegressor(max_depth = 16, random_state = 42)

# Train model

dtr.fit(X_train, y_train)

# Model prediction

dtr_pred = dtr.predict(X_test)

# Error metric

print("Ridge-r2 score:", r2_score(y_test, dtr_pred))

print("Ridge-RMSE:", np.sqrt(mean_squared_error(y_test, dtr_pred)))

# Plot

plt.bar(X_train.columns, dtr.feature_importances_)

plot_tree(dtr);

Decision tree classifier

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

from sklearn.tree import DecisionTreeClassifier

from sklearn.model_selection import train_test_split

from sklearn.metrics import confusion_matrix, classification_report

from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score

from sklearn.tree import plot_tree

# Data load

from sklearn.datasets import load_iris

iris = load_iris()

# Data separation

X = pd.DataFrame(iris['data'], columns = iris['feature_names'])

y = iris['target']

# Split train : test data set = 7:3 and set random seed

X_train, X_test, y_train, y_test = train_test_split(X,

                                                    y,

                                                    test_size = 0.3,

                                                    random_state = 42,

stratify = y)

# Define a model structure

dtc = DecisionTreeClassifier(max_depth = 3, random_state = 42)

# Train model

dtc.fit(X_train, y_train)

# Model prediction

dtc_pred = dtc.predict(X_test)

# Error metric

print(confusion_matrix(y_test, dtc_pred))

print(classification_report(y_test, dtc_pred))

# Confirm coefficient

dtc.feature_importances_

# Plot

plt.bar(X_train.columns, dtc.feature_importances_)

plt.figure(figsize=(8, 8))

plot_tree(dtc,

         feature_names = iris['feature_names'],

         class_names = iris['target_names'],

         filled = True

);

Rolynominal regression + Lasso regression (다항 회귀)

 Ref. : https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.PolynomialFeatures.html

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

from sklearn.preprocessing import PolynomialFeatures

from sklearn.linear_model import Lasso

from sklearn.model_selection import train_test_split

from sklearn.metrics import r2_score, mean_squared_error

# Data load

df = pd.read_csv('./data/xxx.csv')

# Target data separation

y = df['y']

X = df.drop('y', axis = 1)

# Split train : test data set = 7:3 and set random seed

X_train, X_test, y_train, y_test = train_test_split(X,

                                                    y,

                                                    test_size = 0.3,

                                                    random_state = 42)

# Define a model structure for a polynomial variable

poly = PolynomialFeatures(degree = 2, include_bias = False)

# Train polynomial variable

poly.fit(X_train)

# Show feature names

poly.get_feature_names_out()

# poly.get_feature_names()

pd.options.display.max_columns = 200

# Change the data frame to the polynomial variables

poly_X_train = pd.DataFrame(poly.transform(X_train),

                            columns = poly.get_feature_names_out())

# poly_X_train = pd.DataFrame(poly.transform(X_train),

#                             columns = poly.get_feature_names())

# Change the test data to the polynomial variables

poly_X_test = poly.transform(X_test)

# Define a model structure

ls = Lasso() # To reduce some low coefficients

# Train model

ls.fit(poly_X_train, y_train)

# Model prediction

ls_pred = ls.predict(poly_X_test)

# Confirm coefficient

ls.coef_

# Error metric

print("Lasso-r2 score:", r2_score(y_test, ls_pred))

print("Lasso-RMSE:", np.sqrt(mean_squared_error(y_test, ls_pred)))

# Plot

plt.figure(figsize=(20, 20))

plt.barh(poly.get_feature_names_out(), ls.coef_)

Ridge regression

Linear least squares with l2 regularization.

Ref.: https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.Ridge.html

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

from sklearn.linear_model import Ridge

from sklearn.model_selection import train_test_split

from sklearn.metrics import r2_score, mean_squared_error

# Data load

df = pd.read_csv('./data/xxx.csv')

# Target data separation

y = df['y']

X = df.drop('y', axis = 1)

# Split train : test data set = 7:3 and set random seed

X_train, X_test, y_train, y_test = train_test_split(X,

                                                    y,

                                                    test_size = 0.3,

                                                    random_state = 42)

# Define model structure

rg = Ridge(alpha = 0.1)

# Train model

rg.fit(X_train, y_train)

# Model prediction

rg_pred = rg.predict(X_test)

# Confirm coefficient

rg.coef_

# Error metric

print("Ridge-r2 score:", r2_score(y_test, rg_pred))

print("Ridge-RMSE:", np.sqrt(mean_squared_error(y_test, rg_pred)))

# Plot

plt.bar(X_train.columns, rg.coef_)

Linear regression (기본 선형회귀 모델)

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

from sklearn.linear_model import LinearRegression

from sklearn.model_selection import train_test_split

from sklearn.metrics import r2_score, mean_squared_error

# Data load

df = pd.read_csv('./data/xxx.csv')

# Target data separation

y = df['y']

X = df.drop('y', axis = 1)

# Split train : test data set = 7:3 and set random seed

X_train, X_test, y_train, y_test = train_test_split(X,

                                                    y,

                                                    test_size = 0.3,

                                                    random_state = 42)

# Define model structure

lr = LinearRegression()

# Train model

lr.fit(X_train, y_train)

# Model prediction

lr_pred = lr.predict(X_test)

# Confirm coefficient

lr.coef_

# Error metric

print("r2 score:", r2_score(y_test, lr_pred))

print("RMSE:", np.sqrt(mean_squared_error(y_test, lr_pred)))

# Plot

plt.scatter(lr_pred, y_test)