게으른 우루루

import pandas as pd

import os

import matplotlib.pyplot as plt

import mglearn

import numpy as np

path = os.path.join(mglearn.datasets.DATA_PATH, 'ram_price.csv')

ram_prices = pd.read_csv(path)

ram_prices = ram_prices.iloc[:, 1:] # iloc: 정수로 columns접근,  index number : 몇번째 인덱스 까지의 형태

plt.semilogy(ram_prices['date'], ram_prices['price'])

plt.xlabel('year', size=15)

plt.ylabel('price ($/Mbyte)', size=15)

plt.show()

from sklearn.tree import DecisionTreeRegressor

from sklearn.linear_model import LinearRegression

data_train = ram_prices[ram_prices['date'] < 2000]

data_test = ram_prices[ram_prices['date'] >= 2000]

x_train = data_train['date'][:, np.newaxis] # train data를 1열로 만듭니다.

y_train = np.log(data_train['price'])

tree = DecisionTreeRegressor().fit(x_train, y_train)

lr = LinearRegression().fit(x_train, y_train)

# test는 모든 데이터에 대해 적용합니다.

x_all = ram_prices['date'].reshape(-1, 1) # x_all를 1열로 만듭니다.

pred_tree = tree.predict(x_all)

price_tree = np.exp(pred_tree) # log값 되돌리기

pred_lr = lr.predict(x_all)

price_lr = np.exp(pred_lr) # log값 되돌리기

plt.semilogy(ram_prices['date'], price_tree, label='tree predict', ls='--', dashes=(2,1))

plt.semilogy(ram_prices['date'], price_lr, label='linear reg. predict', ls=':')

plt.semilogy(data_train['date'], data_train['price'], label='train data', alpha=0.4)

plt.semilogy(data_test['date'], data_test['price'], label='test data')

plt.legend(loc=1)

plt.xlabel('year', size=15)

plt.ylabel('price ($/Mbyte)', size=15)

plt.show()

ram 가격 데이터로 만든 linear model과 regression tree의 예측값 비교

from mglearn.plots import plot_animal_tree

import matplotlib.pyplot as plt

plot_animal_tree()

plt.show()

몇 가지 동물들을 구분하기 위한 결정 트리

from sklearn.datasets import load_wine

from sklearn.model_selection import train_test_split

from sklearn.tree import DecisionTreeClassifier

wine = load_wine()

x_train, x_test, y_train, y_test = train_test_split(wine.data, wine.target,

                                                    stratify=wine.target, random_state=0)

tree = DecisionTreeClassifier(random_state=0)

tree.fit(x_train, y_train)

score_tr = tree.score(x_train, y_train)

score_te = tree.score(x_test, y_test)

print('{:.3f}'.format(score_tr))

# 1.000

print('{:.3f}'.format(score_te))

# 0.889

tree = DecisionTreeClassifier(max_depth=2, random_state=0)

tree.fit(x_train, y_train)

score_tr = tree.score(x_train, y_train)

score_te = tree.score(x_test, y_test)

print('{:.3f}'.format(score_tr))

# 0.910

print('{:.3f}'.format(score_te))

# 756

import graphviz

from sklearn.tree import export_graphviz

export_graphviz(tree, out_file='tree.dot',

                class_names=wine.target_names,

                feature_names=wine.feature_names,

                impurity=False, # gini 미출력

                filled=True) # filled: node의 색깔을 다르게

with open('tree.dot') as file_reader:

    dot_graph = file_reader.read()

dot = graphviz.Source(dot_graph) # dot_graph의 source 저장

dot.render(filename='tree.png') # png로 저장

tree.png

wine 데이터셋으로 만든 Decision Tree

feature_imp = tree.feature_importances_

print('{}'.format(feature_imp))

# [0.         0.         0.         0.02885475 0.         0.

#  0.         0.         0.         0.43979596 0.         0.

#  0.53134929]

import numpy as np

n_feature = wine.data.shape[1]

idx = np.arange(n_feature)

plt.barh(idx, feature_imp, align='center')

plt.yticks(idx, wine.feature_names)

plt.xlabel('feature importance', size=15)

plt.ylabel('feature', size=15)

plt.show()

feature importance와 feature