通過５折交叉驗證，實現邏輯迴歸，決策樹，SVM,隨機森林，GBDT,Xgboost,lightGBM的評分

• 匯入的包

import pandas as pd
import warnings
from sklearn.preprocessing import scale
from sklearn.model_selection import cross_val_score
from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.svm import SVC
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import GradientBoostingClassifier
from xgboost.sklearn import XGBClassifier
import lightgbm as lgb
 

• 讀取資料集

data_all = pd.read_csv('/home/infisa/wjht/project/DataWhale/data_all.csv', encoding='gbk')

• 處理資料集

df_y=data_all['status']
df_X=data_all.drop(columns=['status'])
df_X=scale(df_X,axis=0)  #將資料轉化為標準資料

• 構建模型

lr = LogisticRegression(random_state=2018,tol=1e-6)  # 邏輯迴歸模型

tree = DecisionTreeClassifier(random_state=2018) #決策樹模型

svm = SVC(probability=True,random_state=2018,tol=1e-6)  # SVM模型

forest=RandomForestClassifier(n_estimators=100,random_state=2018) #　隨機森林

Gbdt=GradientBoostingClassifier(random_state=2018) #CBDT

Xgbc=XGBClassifier(random_state=2018)  #Xgbc

gbm=lgb.LGBMClassifier(random_state=2018)  #lgb
 

• 構建評分函式，並採取５折交叉驗證的方式評分

def muti_score(model):
    warnings.filterwarnings('ignore')
    accuracy = cross_val_score(model, df_X, df_y, scoring='accuracy', cv=5)
    precision = cross_val_score(model, df_X, df_y, scoring='precision', cv=5)
    recall = cross_val_score(model, df_X, df_y, scoring='recall', cv=5)
    f1_score = cross_val_score(model, df_X, df_y, scoring='f1', cv=5)
    auc = cross_val_score(model, df_X, df_y, scoring='roc_auc', cv=5)
    print("準確率:",accuracy.mean())
    print("精確率:",precision.mean())
    print("召回率:",recall.mean())
    print("F1_score:",f1_score.mean())
    print("AUC:",auc.mean())
 

其中mean()指的是求得的均值

• 分析
  模型的評分思想，是通過採用５折交叉驗證，得出其中的均值分數來評判。從上表中可以看出邏輯迴歸，隨機森林，GBDT，Xgboost,LightGbm的各個指標都很相近而且分數也較高，說明這幾個模型擬合數據效果都較好，都可以選做模型。綜合來看Xgboost的分數更高一些，它的評分效果最好。

• 問題
  01 還沒有學會在程式碼中，可以直接輸出表格的操作。
  02 對各個模型的引數還不太瞭解
  03 對資料集劃分，怎樣構造優質的資料還比較欠缺。

• 參考文章
  cross_val_score的 scoring引數值解析
  python機器學習庫sklearn——交叉驗證（K折、留一、留p、隨機）
  12號同學寫的部落格

• 完整程式碼

import pandas as pd
import warnings
from sklearn.preprocessing import scale
from sklearn.model_selection import cross_val_score
from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.svm import SVC
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import GradientBoostingClassifier
from xgboost.sklearn import XGBClassifier
import lightgbm as lgb


# 讀取資料集
data_all = pd.read_csv('/home/infisa/wjht/project/DataWhale/data_all.csv', encoding='gbk')

# 劃分為5折交叉驗證資料集
df_y=data_all['status']
df_X=data_all.drop(columns=['status'])
df_X=scale(df_X,axis=0)  #將資料轉化為標準資料
#構建模型

lr = LogisticRegression(random_state=2018,tol=1e-6)  # 邏輯迴歸模型

tree = DecisionTreeClassifier(random_state=2018) #決策樹模型

svm = SVC(probability=True,random_state=2018,tol=1e-6)  # SVM模型

forest=RandomForestClassifier(n_estimators=100,random_state=2018) #　隨機森林

Gbdt=GradientBoostingClassifier(random_state=2018) #CBDT

Xgbc=XGBClassifier(random_state=2018)  #Xgbc

gbm=lgb.LGBMClassifier(random_state=2018)  #lgb



def muti_score(model):
    warnings.filterwarnings('ignore')
    accuracy = cross_val_score(model, df_X, df_y, scoring='accuracy', cv=5)
    precision = cross_val_score(model, df_X, df_y, scoring='precision', cv=5)
    recall = cross_val_score(model, df_X, df_y, scoring='recall', cv=5)
    f1_score = cross_val_score(model, df_X, df_y, scoring='f1', cv=5)
    auc = cross_val_score(model, df_X, df_y, scoring='roc_auc', cv=5)
    print("準確率:",accuracy.mean())
    print("精確率:",precision.mean())
    print("召回率:",recall.mean())
    print("F1_score:",f1_score.mean())
    print("AUC:",auc.mean())



model_name=["lr","tree","svm","forest","Gbdt","Xgbc","gbm"]
for name in model_name:
    model=eval(name)
    print(name)
    muti_score(model)


'''
lr
準確率: 0.7890191148682617
精確率: 0.6542724662896913
召回率: 0.3377975457965613
F1_score: 0.44525012166067884
AUC: 0.7840451024530857
tree
準確率: 0.6962524533638791
精確率: 0.39920670173446693
召回率: 0.4157413593052284
F1_score: 0.40705496051057793
AUC: 0.6029856787858856
svm
準確率: 0.787758390223099
精確率: 0.7351623295760905
召回率: 0.24060335431243626
F1_score: 0.36179547264664874
AUC: 0.7640376541388867
forest
準確率: 0.7921756804332226
精確率: 0.7135700690071172
召回率: 0.2867128441334693
F1_score: 0.40835414886475174
AUC: 0.7752164698827589
Gbdt
準確率: 0.7938590063951863
精確率: 0.6604108594441386
召回率: 0.36633732991104395
F1_score: 0.4708811551285791
AUC: 0.7888240065764295
Xgbc
準確率: 0.7982740847293591
精確率: 0.6829783239831001
召回率: 0.3663162336064133
F1_score: 0.47673826685376613
AUC: 0.7914190511145234
gbm
準確率: 0.79049080811139
精確率: 0.6421783397519263
召回率: 0.3730354066312717
F1_score: 0.47150438344663004
AUC: 0.7776116341798183
'''