算法实践day2

数据

和day01中的数据一样data_all.csv

任务

使用之前的数据data_all.csv利用随机森林、GBDT、XGBoost和LightGBM这4个模型，评分方式任意。

代码实现

随机森林

1.导入包

import pandas as pd
from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier

2.加载数据

 file_path = 'G:\DatawhaleWeek01\Data\data_all.csv'
 row_data = pd.read_csv(file_path)

3.数据预处理

feature = [x for x in row_data.columns if x not in ['status']]
X = row_data[feature]
# 'status'列是标签
y = row_data['status']
X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.3,random_state=2018)

4.构建模型

rfc = RandomForestClassifier()
rfc.fit(X_train,y_train)
rfc_y_predict = rfc.predict(X_test)
print('使用随机森林的测试集的预测值：', rfc_y_predict)

5.模型评分

rfc_score = rfc.score(X_test,y_test)
y_predict_proba = rfc.predict_proba(X_test)
print('使用随机森林的准确度：', rfc_score)
auc_score = roc_auc_score(y_test,y_predict_proba[:,1])
print('auc的值：', auc_score)
clf_rp = classification_report(y_test, rfc_y_predict, target_names=['Yes','No'])
print('分类结果报表：', clf_rp)

GBDT

1.导入包

 from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier

2.加载数据

 file_path = 'G:\DatawhaleWeek01\Data\data_all.csv'
 row_data = pd.read_csv(file_path)

3.数据预处理

feature = [x for x in row_data.columns if x not in ['status']]
X = row_data[feature]
# 'status'列是标签
y = row_data['status']
X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.3,random_state=2018)

4.构建模型

gbdt = GradientBoostingClassifier(random_state=2018)
gbdt.fit(X_train,y_train)
gbdt_y_predict = gbdt.predict(X_test)
gbdt_y_predict_proba = gbdt.predict_proba(X_test)
print('使用GBDT的测试集的预测值：', gbdt_y_predict)

5.模型评分

gbdt_accuracy_score = gbdt.score(X_test,y_test)
print('使用GBDT的准确度：',gbdt_accuracy_score)
gbdt_auc = roc_auc_score(y_test, gbdt_y_predict_proba[:,1])
print('使用GBDT的auc:',gbdt_auc)

xgboost

1.导入包

 from xgboost.sklearn import XGBClassifier

2.加载数据

 file_path = 'G:\DatawhaleWeek01\Data\data_all.csv'
 row_data = pd.read_csv(file_path)

3.数据预处理

X = row_data.drop(columns=['status'])
y = row_data['status']
X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.3,random_state=2018)

4.构建模型

xgbs = XGBClassifier(random_state=2018)
xgbs.fit(X_train, y_train)
y_predict = xgbs.predict(X_test)
y_predict_proba = xgbs.predict_proba(X_test)
print('预测值：',y_predict)
print('预测值概率：', y_predict_proba)

5.模型评分

y_accuracy_score = accuracy_score(y_test, y_predict)
print('使用XGBOOT的准确率：', y_accuracy_score)
y_auc_score = roc_auc_score(y_test, y_predict_proba[:, 1])
print('使用XGBOOST的auc值：', y_auc_score)

Lightgbm

1.导入包

 import lightgbm as lgb

2.加载数据

 file_path = 'G:\DatawhaleWeek01\Data\data_all.csv'
 row_data = pd.read_csv(file_path)

3.数据预处理

X = row_data.drop(columns=['status'])
y = row_data['status']
X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.3,random_state=2018)
#**转换到lgb的标准数据格式**
lgb_train = lgb.Dataset(data=X_train, label=y_train)
lgb_test = lgb.Dataset(data=X_test, label=y_test)

4.构建模型

# 训练lgb分类器
params = {
        'task': 'train',
        'objective': 'binary',
        'learning_rate': 0.1,
        'num_leaves':31,
        'max_depth':-1,
        'lambda_l1': 0,
        'lambda_l2': 0.5,
        'bagging_fraction' :1.0,
        'feature_fraction': 1.0
        }

bst = lgb.train(params,lgb_train,num_boost_round=800,valid_sets=lgb_test, early_stopping_rounds=5)

5.模型评分

# 得到预测概率（测试集只能使用原生数据集，不能使用lgb标准化之后的数据）
y_predict_proba = bst.predict(X_test, num_iteration=bst.best_iteration)
print(y_predict_proba)
y_accuracy_score = accuracy_score(y_test, y_predict)
print('使用XGBOOT的准确率：', y_accuracy_score)
y_auc_score = roc_auc_score(y_test, y_predict_proba[:, 1])
print('使用XGBOOST的auc值：', y_auc_score)