【阿旭机器学习实战】【35】员工离职率预测---决策树与随机森林预测

# 引入工具包
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib as matplot
import seaborn as sns
%matplotlib inline

# 读入数据到Pandas Dataframe "df"
df = pd.read_csv('HR_comma_sep.csv', index_col=None)

# 检测是否有缺失数据
df.isnull().any()

satisfaction_level       False
last_evaluation          False
number_project           False
average_montly_hours     False
time_spend_company       False
Work_accident            False
left                     False
promotion_last_5years    False
sales                    False
salary                   False
dtype: bool

# 数据的样例
df.head()

# 重命名
df = df.rename(columns={'satisfaction_level': 'satisfaction', 
                        'last_evaluation': 'evaluation',
                        'number_project': 'projectCount',
                        'average_montly_hours': 'averageMonthlyHours',
                        'time_spend_company': 'yearsAtCompany',
                        'Work_accident': 'workAccident',
                        'promotion_last_5years': 'promotion',
                        'sales' : 'department',
                        'left' : 'turnover'
                        })

# 将预测标签‘是否离职’放在第一列
front = df['turnover']
df.drop(labels=['turnover'], axis=1, inplace = True)
df.insert(0, 'turnover', front)
df.head()

df.shape

(14999, 10)

# 特征数据类型. 
df.dtypes

turnover                 int64
satisfaction           float64
evaluation             float64
projectCount             int64
averageMonthlyHours      int64
yearsAtCompany           int64
workAccident             int64
promotion                int64
department              object
salary                  object
dtype: object

turnover_rate = df.turnover.value_counts() / len(df)
turnover_rate

0    0.761917
1    0.238083
Name: turnover, dtype: float64

# 显示统计数据
df.describe()

# 分组的平均数据统计
turnover_Summary = df.groupby('turnover')
turnover_Summary.mean()

# 相关性矩阵
corr = df.corr()
#corr = (corr)
sns.heatmap(corr, 
            xticklabels=corr.columns.values,
            yticklabels=corr.columns.values)
corr

# 比较离职和未离职员工的满意度
emp_population = df['satisfaction'][df['turnover'] == 0].mean()
emp_turnover_satisfaction = df[df['turnover']==1]['satisfaction'].mean()
print( '未离职员工满意度: ' + str(emp_population))
print( '离职员工满意度: ' + str(emp_turnover_satisfaction) )

未离职员工满意度: 0.666809590479516
离职员工满意度: 0.44009801176140917

import scipy.stats as stats
stats.ttest_1samp(a = df[df['turnover']==1]['satisfaction'], # 离职员工的满意度样本
                  popmean = emp_population)  # 未离职员工的满意度均值

Ttest_1sampResult(statistic=-51.3303486754725, pvalue=0.0)

degree_freedom = len(df[df['turnover']==1])
LQ = stats.t.ppf(0.025,degree_freedom)  # 95%致信区间的左边界
RQ = stats.t.ppf(0.975,degree_freedom)  # 95%致信区间的右边界
print ('The t-分布 左边界: ' + str(LQ))
print ('The t-分布 右边界: ' + str(RQ))

The t-分布 左边界: -1.9606285215955626
The t-分布 右边界: 1.9606285215955621

# 概率密度函数估计
fig = plt.figure(figsize=(15,4),)
ax=sns.kdeplot(df.loc[(df['turnover'] == 0),'evaluation'] , color='b',shade=True,label='no turnover')
ax=sns.kdeplot(df.loc[(df['turnover'] == 1),'evaluation'] , color='r',shade=True, label='turnover')
ax.set(xlabel='Employee Evaluation', ylabel='Frequency')
ax.legend()
plt.title('Employee Evaluation Distribution - Turnover V.S. No Turnover')

Text(0.5, 1.0, 'Employee Evaluation Distribution - Turnover V.S. No Turnover')

# 概率密度函数估计
fig = plt.figure(figsize=(15,4))
ax=sns.kdeplot(df.loc[(df['turnover'] == 0),'averageMonthlyHours'] , color='b',shade=True, label='no turnover')
ax=sns.kdeplot(df.loc[(df['turnover'] == 1),'averageMonthlyHours'] , color='r',shade=True, label='turnover')
ax.legend()
ax.set(xlabel='Employee Average Monthly Hours', ylabel='Frequency')
plt.title('Employee AverageMonthly Hours Distribution - Turnover V.S. No Turnover')

Text(0.5, 1.0, 'Employee AverageMonthly Hours Distribution - Turnover V.S. No Turnover')

# 概率密度函数估计
fig = plt.figure(figsize=(15,4))
ax=sns.kdeplot(df.loc[(df['turnover'] == 0),'satisfaction'] , color='b',shade=True, label='no turnover')
ax=sns.kdeplot(df.loc[(df['turnover'] == 1),'satisfaction'] , color='r',shade=True, label='turnover')
plt.title('Employee Satisfaction Distribution - Turnover V.S. No Turnover')
ax.legend()

<matplotlib.legend.Legend at 0x281a5a6b820>

from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, classification_report, precision_score, recall_score, confusion_matrix, precision_recall_curve

# 将string类型转换为整数类型
df["department"] = df["department"].astype('category').cat.codes
df["salary"] = df["salary"].astype('category').cat.codes
# 产生X, y
target_name = 'turnover'
X = df.drop('turnover', axis=1)
y = df[target_name]
# 将数据分为训练和测试数据集
# 注意参数 stratify = y 意味着在产生训练和测试数据中, 离职的员工的百分比等于原来总的数据中的离职的员工的百分比
X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.15, random_state=123, stratify=y)
df.head()

from sklearn.metrics import roc_auc_score
from sklearn.metrics import classification_report
from sklearn.ensemble import RandomForestClassifier
from sklearn import tree
from sklearn.tree import DecisionTreeClassifier
# 决策树
dtree = tree.DecisionTreeClassifier(
    criterion='entropy',
    #max_depth=3, # 定义树的深度, 可以用来防止过拟合
    min_weight_fraction_leaf=0.01 # 定义叶子节点最少需要包含多少个样本(使用百分比表达), 防止过拟合
    )
dtree = dtree.fit(X_train,y_train)
print ("\n\n ---决策树---")
dt_roc_auc = roc_auc_score(y_test, dtree.predict(X_test))
print ("决策树 AUC = %2.2f" % dt_roc_auc)
print(classification_report(y_test, dtree.predict(X_test)))
# 随机森林
rf = RandomForestClassifier(
    criterion='entropy',
    n_estimators=1000, 
    max_depth=None, # 定义树的深度, 可以用来防止过拟合
    min_samples_split=10, # 定义至少多少个样本的情况下才继续分叉
    #min_weight_fraction_leaf=0.02 # 定义叶子节点最少需要包含多少个样本(使用百分比表达), 防止过拟合
    )
rf.fit(X_train, y_train)
print ("\n\n ---随机森林---")
rf_roc_auc = roc_auc_score(y_test, rf.predict(X_test))
print ("随机森林 AUC = %2.2f" % rf_roc_auc)
print(classification_report(y_test, rf.predict(X_test)))

---决策树---
决策树 AUC = 0.93
              precision    recall  f1-score   support
           0       0.97      0.98      0.97      1714
           1       0.93      0.89      0.91       536
    accuracy                           0.96      2250
   macro avg       0.95      0.93      0.94      2250
weighted avg       0.96      0.96      0.96      2250
 ---随机森林---
随机森林 AUC = 0.97
              precision    recall  f1-score   support
           0       0.98      1.00      0.99      1714
           1       0.99      0.94      0.97       536
    accuracy                           0.98      2250
   macro avg       0.99      0.97      0.98      2250
weighted avg       0.98      0.98      0.98      2250

# ROC 图
from sklearn.metrics import roc_curve
rf_fpr, rf_tpr, rf_thresholds = roc_curve(y_test, rf.predict_proba(X_test)[:,1])
dt_fpr, dt_tpr, dt_thresholds = roc_curve(y_test, dtree.predict_proba(X_test)[:,1])
plt.figure()
# 随机森林 ROC
plt.plot(rf_fpr, rf_tpr, label='Random Forest (area = %0.2f)' % rf_roc_auc)
# 决策树 ROC
plt.plot(dt_fpr, dt_tpr, label='Decision Tree (area = %0.2f)' % dt_roc_auc)
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.title('ROC Graph')
plt.legend(loc="lower right")
plt.show()

## 画出决策树特征的重要性 ##
importances = rf.feature_importances_
feat_names = df.drop(['turnover'],axis=1).columns
indices = np.argsort(importances)[::-1]
plt.figure(figsize=(12,6))
plt.title("Feature importances by RandomForest")
plt.bar(range(len(indices)), importances[indices], color='lightblue',  align="center")
plt.step(range(len(indices)), np.cumsum(importances[indices]), where='mid', label='Cumulative')
plt.xticks(range(len(indices)), feat_names[indices], rotation='vertical',fontsize=14)
plt.xlim([-1, len(indices)])
plt.show()

## 画出决策树的特征的重要性 ##
importances = dtree.feature_importances_
feat_names = df.drop(['turnover'],axis=1).columns
indices = np.argsort(importances)[::-1]
plt.figure(figsize=(12,6))
plt.title("Feature importances by Decision Tree")
plt.bar(range(len(indices)), importances[indices], color='lightblue',  align="center")
plt.step(range(len(indices)), np.cumsum(importances[indices]), where='mid', label='Cumulative')
plt.xticks(range(len(indices)), feat_names[indices], rotation='vertical',fontsize=14)
plt.xlim([-1, len(indices)])
plt.show()