数据分析实战—玻璃类别分类

1.实战内容

(1) 加载玻璃类别数据集，划分训练集、测试集

import pandas as pd
import numpy as np
pd.set_option('display.max_columns', None)
pd.set_option('display.max_rows', None)
# 读取数据集
glass = pd.read_csv('glass.csv', encoding='gbk', header=None)
glass_data = glass.iloc[1:, : -1]
glass_target = glass.iloc[1:, -1]
glass

glass_data

glass_target

# 划分训练集和测试集
from sklearn.model_selection import train_test_split
glass_data_train, glass_data_test, \
glass_target_train, glass_target_test = \
train_test_split(glass_data, glass_target, test_size=0.2, random_state=6)

glass_data_train

glass_data_test

(2) 对训练集、测试集进行标准差标准化，并分别输出标准化之后的训练集、测试集的方差和均值。

# 标准差标准化
from sklearn.preprocessing import StandardScaler
stdScale = StandardScaler().fit(glass_data_train)
glass_trainScaler = stdScale.transform(glass_data_train)
glass_testScaler = stdScale.transform(glass_data_test)
print('标准差标准化后训练集数据的方差为：', np.var(glass_trainScaler))
print('标准差标准化后训练集数据的均值为：', np.mean(glass_trainScaler))
print('标准差标准化后测试集数据的方差为：', np.var(glass_testScaler))
print('标准差标准化后测试集数据的均值为：', np.mean(glass_testScaler))

glass_trainScaler

(3) 使用支持向量机对玻璃类别数据集进行分类，输出分类模型评价报告。

# PCA降维
from sklearn.decomposition import PCA
pca = PCA(n_components=0.999).fit(glass_data_train)
glass_trainPca = pca.transform(glass_data_train)
glass_testPca = pca.transform(glass_data_test)

# SVM分类模型
from sklearn.svm import SVC
svm = SVC().fit(glass_trainScaler, glass_target_train)
print('建立的SVM模型为：\n', svm)
glass_target_pred = svm.predict(glass_testScaler)
print('预测前10个结果为：\n', glass_target_pred[: 10])
# 打印出分类报告,评价分类模型性能
from sklearn.metrics import classification_report
print('使用SVM预测glass数据的分类报告为：', '\n',
classification_report(glass_target_test, glass_target_pred))

(4) 使用梯度提升回归树对玻璃类别数据集进行回归，并计算回归模型的5项评价指标得分。

from sklearn.ensemble import GradientBoostingRegressor
GBR_glass = GradientBoostingRegressor().fit(glass_trainPca, glass_target_train)
glass_target_pred = GBR_glass.predict(glass_testPca)
# 回归模型评价指标
from sklearn.metrics import explained_variance_score, mean_absolute_error, mean_squared_error, median_absolute_error, r2_score
print('glass数据梯度提升回归树模型的平均绝对误差为：',
mean_absolute_error(glass_target_test, glass_target_pred))
print('glass数据梯度提升回归树模型的均方误差为：',
mean_squared_error(glass_target_test, glass_target_pred))
print('glass数据梯度提升回归树模型的中值绝对误差为：',
median_absolute_error(glass_target_test, glass_target_pred))
print('glass数据梯度提升回归树模型的可解释方差值为：',
explained_variance_score(glass_target_test, glass_target_pred))
print('glass数据梯度提升回归树模型的 R 方值为：',
r2_score(glass_target_test, glass_target_pred))