数据分析-分类-案例

目录

1、导入必要的包

2、读入数据

3、数据探索

4、数据预处理

6、建立模型

6.1逻辑回归

6.2 K近邻

6.3 决策树

6.4 朴素贝叶斯

6.5 SVC

---

---

0、数据集介绍

数据集来源：UCI

UCI Machine Learning Repository: Car Evaluation Data Set

1. Title: Car Evaluation Database

2. Sources:
   (a) Creator: Marko Bohanec
   (b) Donors: Marko Bohanec (marko.bohanec@ijs.si)
   Blaz Zupan (blaz.zupan@ijs.si)
   (c) Date: June, 1997

3. Number of Instances: 1728

4. Attribute Values:

   buying v-high, high, med, low
   maint v-high, high, med, low
   doors 2, 3, 4, 5-more
   persons 2, 4, more
   lug_boot small, med, big
   safety low, med, high

5. Missing Attribute Values: none

6. Class Distribution (number of instances per class)

        unacc 1210 (70.023 %)
        acc 384 (22.222 %)
        good 69 ( 3.993 %)
        v-good 65 ( 3.762 %)

1、导入必要的包

from sklearn.linear_model import LogisticRegression
from sklearn.metrics import classification_report
import matplotlib.pyplot as plt
import pandas as pd

2、读入数据

读入数据，并为每个特征命名。

car_data=pd.read_csv("d:/datasets/car_evaluation.csv",names=["buying","maint","doors","person","lug_boot","safety","car_acc"])

3、数据探索

car_data.head()
car_data.tail()
car_data.describe()
car_data.info()
#查看特征与标签的取值
for i in car_data.columns:
    print(car_data[i].value_counts())

4、数据预处理

标签编码（map方法）。

car_data.car_acc=car_data.car_acc.map({"unacc":1,"acc":0,"good":0,"vgood":0})

特征编码get_dummies()）

car_data_=pd.get_dummies(car_data[["lug_boot","safety","buying","maint","doors","person"]])

5、训练集、测试集划分

train_x,test_x,train_y,test_y=train_test_split(car_data_,car_data.car_acc,test_size=0.3,random_state=10)

6、建立模型

6.1逻辑回归

建立模型、训练、预测

lr=LogisticRegression()
lr.fit(train_x,train_y)
y_pre=lr.predict(test_x)

评价

print(classification_report(y_pre,test_y))  #输出分类报告
print(lr.coef_)   #输出回归系数
lr.score(test_x,test_y)  #输出正确率
from sklearn.metrics import accuracy_score
accuracy_score(y_pre,test_y) #输出正确率

from  sklearn.metrics import confusion_matrix
import seaborn as sns
sns.set(font="SimHei")
lgr=LogisticRegression(random_state=1)
lgr.fit(X_train,Y_train)
y_pre=lgr.predict(x_test)
print(lgr.score(x_test,y_test))
ax=sns.heatmap(confusion_matrix(y_test,y_pre),annot=True,fmt="d",xticklabels=["满意","不满意"],yticklabels=["满意","不满意"])
ax.set_ylabel("真实")
ax.set_xlabel("预测")
ax.set_title("混淆矩阵")

6.2 K近邻

建立模型、训练、预测

from sklearn.neighbors import KNeighborsClassifier
kn=KNeighborsClassifier()
kn.fit(X_train,Y_train)
y_pre=kn.predict(x_test)

评价

print(kn.score(x_test,y_test))
print(classification_report(y_test,y_pre))

ax=sns.heatmap(confusion_matrix(y_test,y_pre),annot=True,fmt="d",xticklabels=["满意","不满意"],yticklabels=["满意","不满意"])
ax.set_ylabel("真实")
ax.set_xlabel("预测")
ax.set_title("混淆矩阵")

调参（不同邻居数量对模型的效果的影响）

k_score=[KNeighborsClassifier(n_neighbors=k).fit(X_train,Y_train).score(x_test,y_test) for k in range(1,10)]

import matplotlib.pyplot as plt
plt.plot(range(1,10),k_score,"b")
plt.xlabel("k")
plt.ylabel("正确率")
plt.title("k值对正确率的影响")

6.3 决策树

建立模型、训练、预测

from sklearn.tree import DecisionTreeClassifier
dt=DecisionTreeClassifier(max_depth=3)  #深度较小，为方便可视化
dt.fit(train_x,train_y)
dt_pre=dt.predict(test_x)

评价

dt.score(test_x,test_y)
print(classification_report(dt_pre,test_y))

可视化决策树

前提：已安装graphviz

安装graphviz过程：

• 下载graphviz，选择与操作系统对应的版本  Download | Graphviz
• 安装graphviz
• 修改环境变量（将dot.exe文件的路径加到用户或系统环境变量的path中，一般是：C:\Program Files\Graphviz\bin） 
• 安装graphviz python包：pip install graphviz
• 重启 jupyter

from sklearn.tree import export_graphviz
from six import StringIO
from IPython.display import Image 
import pydotplus
plt.rcParams['font.sans-serif']=['SimHei']  #正常显示中文
plt.rcParams['axes.unicode_minus']=False   #正常显示负号“-”

## 输出图片到dot文件
export_graphviz(dt, out_file='tree.dot', 
                feature_names=train_x.columns,
                rounded=True, filled=True,
                class_names=['acc', 'unacc'])
## 使用dot文件构造图
graph= pydotplus.graph_from_dot_file('tree.dot')
Image(graph.create_png())

6.4 朴素贝叶斯

from sklearn.naive_bayes import GaussianNB  
from sklearn.naive_bayes import BernoulliNB
from sklearn.naive_bayes import MultinomialNB

GNB=GaussianNB()
GNB.fit(train_x,train_y)
y_pred=GNB.predict(test_x)
GNB.score(test_x,test_y)
print(classification_report(test_y,y_pred))

pd.DataFrame(GNB.theta_.T,index=test_x.columns).rename(columns={0:"0类",1:"1类"}).plot(kind="bar",rot=60,figsize=(12,5))

BNB=BernoulliNB()
BNB.fit(train_x,train_y)
y_pred=BNB.predict(test_x)
BNB.score(test_x,test_y)
print(classification_report(test_y,y_pred))

MNB=MultinomialNB()
MNB.fit(train_x,train_y)
y_pred=MNB.predict(test_x)
MNB.score(test_x,test_y)
print(classification_report(test_y,y_pred))

6.5 SVC

from sklearn.svm import SVC
from sklearn.svm import LinearSVC

SVC_=SVC()
SVC_.fit(train_x,train_y)
y_pred=SVC_.predict(test_x)
SVC_.score(test_x,test_y)
print(classification_report(test_y,y_pred))

LSVC_=LinearSVC()
LSVC_.fit(train_x,train_y)
y_pred=LSVC_.predict(test_x)
LSVC_.score(test_x,test_y)
print(classification_report(test_y,y_pred))