美国收入预测Adult数据集决策树分类

"""
该脚本创建分类树实现美国收入的预测
重点是采用pandas实现对数据的处理（非数值数据的转换）以及混淆矩阵的绘制
从决策树到随机森林https://zhuanlan.zhihu.com/p/28217071
"""

import pandas as pd
import numpy as np
# from plotnine import *
import matplotlib.pyplot as plt
from sklearn.preprocessing import LabelEncoder
from sklearn_pandas import DataFrameMapper
# https://www.jianshu.com/p/29af03788ff6 使用pip安装
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier

# 数据集从http://archive.ics.uci.edu/ml/datasets/Adult 下载
下载后数据
training_data = 'adult.data'
test_data = 'adult.test'
columns = ['Age','Workclass','fnlgwt','Education','EdNum','MaritalStatus',
           'Occupation','Relationship','Race','Sex','CapitalGain',
           'CapitalLoss','HoursPerWeek','Country','Income']
df_train_set = pd.read_csv(training_data, names=columns)
df_test_set = pd.read_csv(test_data, names=columns, skiprows=1) # skiprows 忽略第一行
# https://blog.csdn.net/weixin_44056331/article/details/89366105

df_train_set.drop('fnlgwt', axis=1, inplace=True) # 删除fnlgwt列
df_test_set.drop('fnlgwt', axis=1, inplace=True)
# =============================================================================
# 数据清洗
#replace the special character to "Unknown"
for i in df_train_set.columns:
    df_train_set[i].replace('?', 'Unknown', inplace=True)
    df_test_set[i].replace('?', 'Unknown', inplace=True)
    for col in df_train_set.columns:
        if df_train_set[col].dtype != 'int64':
            df_train_set[col] = df_train_set[col].apply(lambda val: val.replace(" ", ""))
            df_train_set[col] = df_train_set[col].apply(lambda val: val.replace(".", ""))
            df_test_set[col] = df_test_set[col].apply(lambda val: val.replace(" ", ""))
            df_test_set[col] = df_test_set[col].apply(lambda val: val.replace(".", ""))
 
# Eduction 和 EdNum特征相似，可以删除Eduction
# Country对收入影响不大，也可以删除
df_train_set.drop(["Country", "Education"], axis=1, inplace=True) 
df_test_set.drop(["Country", "Education"], axis=1, inplace=True)

# Age 和 EdNum 列是数值型的，将连续数值型转换为更高效的方式，
# 比如将年龄转换为10的整数倍，教育年限转换为5的整数倍
colnames = list(df_train_set.columns) # 将数据的表头转换为列表并储存
colnames.remove('Age')
colnames.remove('EdNum')
colnames = ['AgeGroup', 'EduGroup']+colnames

labels = ["{0}-{1}".format(i, i+9) for i in range(0,100,10)]
# https://www.runoob.com/python/att-string-format.html 利用format快速创建字符型列表
df_train_set['AgeGroup'] = pd.cut(df_train_set.Age, range(0,101,10), right = False, labels = labels)
# https://www.cnblogs.com/sench/p/10128216.html 用来把一组数据分割成离散的区间 
df_test_set['AgeGroup'] = pd.cut(df_test_set.Age, range(0,101,10), right = False, labels = labels) 

labels = ["{0}-{1}".format(i,i+4) for i in range(0,20,5)]    
df_train_set['EduGroup'] = pd.cut(df_train_set.EdNum, range(0,21,5), right = False, labels = labels)   
df_test_set['EduGroup'] = pd.cut(df_test_set.EdNum, range(0,21,5), right = False, labels = labels)

df_train_set = df_train_set[colnames] # 仅提取包含colnames中的列，且按照colnames排序
df_test_set = df_test_set[colnames]
# =============================================================================
# 将非数值型数据转换为数值型数据
# df_train_set.Income.value_counts()
# df_test_set.Income.value_counts()
mapper = DataFrameMapper([('AgeGroup', LabelEncoder()),('EduGroup', LabelEncoder()),
                          ('Workclass', LabelEncoder()),('MaritalStatus', LabelEncoder()),
                          ('Occupation', LabelEncoder()),('Relationship', LabelEncoder()),
                          ('Race', LabelEncoder()),('Sex', LabelEncoder()),
                          ('Income', LabelEncoder())], df_out=True, default=None)

cols = list(df_train_set.columns)
cols.remove('Income')
cols = cols[:-3]+['Income']+cols[-3:]  # 将Income列转移到中间去

df_train = mapper.fit_transform(df_train_set.copy())
df_train.columns = cols

df_test = mapper.transform(df_test_set.copy())
df_test.columns = cols

cols.remove('Income')

# 训练数据与测试数据划分
x_train, y_train = df_train[cols].values, df_train['Income'].values
x_test, y_test = df_test[cols].values, df_test['Income'].values
# =============================================================================
# 模型初步训练与评分
treeClassifier = DecisionTreeClassifier()
treeClassifier.fit(x_train, y_train)
score = treeClassifier.score(x_test, y_test)
print('The score before GridSearch:', score)
# ===========================================================================
# Plot confusion matrix.
from sklearn.metrics import plot_confusion_matrix
np.set_printoptions(precision=4)
titles_options = [('Confusion matrix withou normalization',None),
                  ('Normalized confusion matrix', 'true')]
class_names = [df_test_set['Income'][1],df_test_set['Income'][2]]
for title, normalize in titles_options:
    disp = plot_confusion_matrix(treeClassifier, x_test, y_test,
                                 display_labels = class_names,
                                 cmap = plt.cm.Oranges,
                                 normalize = normalize)
    disp.ax_.set_title(title)
    print(title)
    print(disp.confusion_matrix)
plt.show()
plt.ylabel('True label')
plt.xlabel('Predicted label')
# =============================================================================
# 采用交叉验证进行参数优化
from sklearn.model_selection import GridSearchCV
paras = {'max_features':(None, 9, 6),
         'max_depth':(None, 24, 16),
         'min_samples_split':(2, 4, 8),
         'min_samples_leaf':(16, 4, 12)
    }

clf = GridSearchCV(treeClassifier, paras, cv = 5)
clf.fit(x_train, y_train)
clf.best_score_, clf.score(x_test, y_test), clf.best_params_
print('The best score after GridSearch:', clf.best_score_)
print('The score after GridSearch:', clf.score(x_test, y_test))
print('The best parameters:', clf.best_params_)
disp2 = plot_confusion_matrix(clf, x_test, y_test,
                              display_labels = class_names,
                              cmap = plt.cm.Blues,
                              normalize = 'true')
disp2.ax_.set_title('Confusion matrix after GridSearch')
print(title)
print(disp2.confusion_matrix)

主要参考资料

从决策树到随机森林
Confusion matrix
分类模型主要评估指标