机器学习：sklearn模型指标和特征贡献度查看

模型训练完成后，即使模型评估很好，各项指标都很到位，业务人员肯定也是心里没底的，哪怕有模型公式，他们也看不懂啊。咋整，当然是先把模型的重要评估指标打印给他们看，再把特征贡献度从大到小，画成图给他们看啦。今天就通过sklearn实现模型评估指标和特征贡献度的图形查看。

本文的数据集采用泰坦尼克号幸存者数据。使用sklearn的决策树和随机森林进行预测，然后查看模型的评估指标，最后将特征的贡献度从大到小以柱状图展示。

直接上代码：

# -*- coding: utf-8 -*-
# @Time    : 2018/12/13 上午10:30
# @Author  : yangchen
# @FileName: featureimportance.py
# @Software: PyCharm
# @Blog    ：https://blog.youkuaiyun.com/opp003/article

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import accuracy_score, precision_recall_fscore_support
from sklearn.model_selection import train_test_split

#导入数据
df = pd.read_csv('processed_titanic.csv', header=0)

#设置y值
X = df.drop(["survived"], axis=1)
y = df["survived"]

#训练集和测试集划分
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=0, shuffle=True)

#训练模型
dtree = DecisionTreeClassifier(criterion="entropy", random_state=123,
                               max_depth=4,
                               min_samples_leaf=5)
dtree.fit(X_train, y_train)

# 预测
pred_train = dtree.predict(X_train)
pred_test = dtree.predict(X_test)

#准确率
train_acc = accuracy_score(y_train, pred_train)
test_acc = accuracy_score(y_test, pred_test)
print ("训练集准确率: {0:.2f}, 测试集准确率: {1:.2f}".format(train_acc, test_acc))

#其他模型评估指标
precision, recall, F1, _ = precision_recall_fscore_support(y_test, pred_test, average="binary")
print ("精准率: {0:.2f}. 召回率: {1:.2f}, F1分数: {2:.2f}".format(precision, recall, F1))

#特征重要度
features = list(X_test.columns)
importances = dtree.feature_importances_
indices = np.argsort(importances)[::-1]
num_features = len(importances)

#将特征重要度以柱状图展示
plt.figure()
plt.title("Feature importances")
plt.bar(range(num_features), importances[indices], color="g", align="center")
plt.xticks(range(num_features), [features[i] for i in indices], rotation='45')
plt.xlim([-1, num_features])
plt.show()

#输出各个特征的重要度
for i in indices:
    print ("{0} - {1:.3f}".format(features[i], importances[i]))

核心代码：

importances = dtree.feature_importances_
indices = np.argsort(importances)[::-1]

第一行，是通过模型的feature_importances_方法获取特征贡献度。使用sklearn的算法都有这个方法。然后使用argsort对其进行排序，由于argsort排序是从小到大的，因此要用[::-1]进行倒序，得到从大到小的排序。

整个代码运行得到输出如下：

训练集准确率: 0.80, 测试集准确率: 0.79
精准率: 1.00. 召回率: 0.70, F1分数: 0.82
sex - 0.543
age - 0.247
fare - 0.211
embarked - 0.000
parch - 0.000
sibsp - 0.000
pclass - 0.000

特征贡献度图形如下：

 

另外，也可以使用score方法获取模型得分，但是参数得换下：

train_acc = dtree.score(X_train, y_train)
test_acc = dtree.score(X_test, y_test)

最后再来看下，使用随机森林的方法，表现如何：

# -*- coding: utf-8 -*-
# @Time    : 2018/12/14 上午09:13
# @Author  : yangchen
# @FileName: randomforests.py
# @Software: PyCharm
# @Blog    ：https://blog.youkuaiyun.com/opp003/article

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score, precision_recall_fscore_support
from sklearn.model_selection import train_test_split
from sklearn import metrics

#导入数据
df = pd.read_csv('processed_titanic.csv', header=0)

#设置y值
X = df.drop(["survived"], axis=1)
y = df["survived"]

#训练集和测试集划分
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=0, shuffle=True)

#训练模型
forest = RandomForestClassifier(
    n_estimators=10, criterion="entropy",
    max_depth=4, min_samples_leaf=5)
forest.fit(X_train, y_train)

# 预测
pred_train = forest.predict(X_train)
pred_test = forest.predict(X_test)

#准确率
train_acc = accuracy_score(y_train, pred_train)
test_acc = accuracy_score(y_test, pred_test)
print ("训练集准确率: {0:.2f}, 测试集准确率: {1:.2f}".format(train_acc, test_acc))

#其他模型评估指标
precision, recall, F1, _ = precision_recall_fscore_support(np.array(y_test), np.array(pred_test), average='binary')


#特征重要度
features = list(X_test.columns)
importances = forest.feature_importances_
indices = np.argsort(importances)[::-1]
num_features = len(importances)

#将特征重要度以柱状图展示
plt.figure()
plt.title("Feature importances")
plt.bar(range(num_features), importances[indices], color="g", align="center")
plt.xticks(range(num_features), [features[i] for i in indices], rotation='45')
plt.xlim([-1, num_features])
plt.show()

#输出各个特征的重要度
for i in indices:
    print ("{0} - {1:.3f}".format(features[i], importances[i]))

输出结果：

训练集准确率: 0.80, 测试集准确率: 0.75
precision: 0.83. recall: 0.81, F1: 0.82
sex - 0.437
age - 0.300
fare - 0.176
pclass - 0.034
parch - 0.031
embarked - 0.016
sibsp - 0.007

特征贡献度图形如下：

 

也可以这么写，来查看特征重要度：

（12-19补充另一种写法，由于数据出了问题，导致图形的不一致，主要看方法。。。）

features = list(X_test.columns)

importance_frame = pd.DataFrame({'Importance': list(dtree.feature_importances_), 'Feature': list(features)})
importance_frame.sort_values(by='Importance', inplace=True)
#importance_frame['rela_imp'] = importance_frame['Importance'] / sum(importance_frame['Importance'])
importance_frame.plot(kind='barh', x='Feature', figsize=(8, 8), color='orange')
plt.show()