阿里天池金融数据分析赛题2：保险反欺诈预测baseline

金融数据分析赛题2：保险反欺诈预测baseline

好久没写baseline了，最近逛比赛的时候突然看到阿里新人赛又出新题目了，索性写个baseline给初学者，昨天晚上把比赛数据下载了，然后随便跑了个模型，AUC就达到了0.95，排在了第二名，下图是我排名的截图，所以题目还是比较简单的，适合初学者入手。

比赛地址：https://tianchi.aliyun.com/competition/entrance/531994/introduction?spm=5176.12281973.1005.21.3dd52448vSKXI0

我比较喜欢做开源，因为分享也是一种快乐，如果大家对baseline代码有任何疑问，都可以提出来，我会详细解答的，也欢迎大家关注，有任何问题我都会解答！

话不多说，直接上代码吧！

baseline代码如下：

import pandas as pd
import datetime
import warnings
warnings.filterwarnings('ignore')
from sklearn.model_selection import StratifiedKFold
#warnings.filterwarnings('ignore')
#%matplotlib inline
from sklearn.metrics import roc_auc_score
## 数据降维处理的
from sklearn.model_selection import train_test_split  
from catboost import CatBoostClassifier
from sklearn.preprocessing import LabelEncoder
train=pd.read_csv("E:/金融2/train.csv")
test=pd.read_csv("E:/金融2/test.csv")
sub=pd.read_csv("E:/金融2/submission.csv")
data=pd.concat([train,test])
data['incident_date'] = pd.to_datetime(data['incident_date'],format='%Y-%m-%d')
startdate = datetime.datetime.strptime('2022-06-30', '%Y-%m-%d')
data['time'] = data['incident_date'].apply(lambda x: startdate-x).dt.days
#Encoder
numerical_fea = list(data.select_dtypes(include=['object']).columns)
division_le = LabelEncoder()
for fea in numerical_fea:
    division_le.fit(data[fea].values)
    data[fea] = division_le.transform(data[fea].values) 
print("数据预处理完成!")
testA=data[data['fraud'].isnull()].drop(['policy_id','incident_date','fraud'],axis=1)
trainA=data[data['fraud'].notnull()]
data_x=trainA.drop(['policy_id','incident_date','fraud'],axis=1)
data_y=train[['fraud']].copy()
col=['policy_state','insured_sex','insured_education_level','incident_type','collision_type','incident_severity','authorities_contacted','incident_state',
     'incident_city','police_report_available','auto_make','auto_model']
for i in data_x.columns:
    if i in col:
        data_x[i] = data_x[i].astype('str')
for i in testA.columns:
    if i in col:
        testA[i] = testA[i].astype('str')       
model=CatBoostClassifier(
            loss_function="Logloss",
            eval_metric="AUC",
            task_type="CPU",
            learning_rate=0.1,
            iterations=10000,
            random_seed=2020,
            od_type="Iter",
            depth=7,
            early_stopping_rounds=300)
answers = []
mean_score = 0
n_folds = 10
sk = StratifiedKFold(n_splits=n_folds, shuffle=True, random_state=2019)
for train, test in sk.split(data_x, data_y):
    x_train = data_x.iloc[train]
    y_train = data_y.iloc[train]
    x_test = data_x.iloc[test]
    y_test = data_y.iloc[test]
    clf = model.fit(x_train,y_train, eval_set=(x_test,y_test),verbose=500,cat_features=col)
    yy_pred_valid=clf.predict(x_test)
    print('cat验证的auc:{}'.format(roc_auc_score(y_test, yy_pred_valid)))
    mean_score += roc_auc_score(y_test, yy_pred_valid) / n_folds
    y_pred_valid = clf.predict(testA,prediction_type='Probability')[:,-1]
    answers.append(y_pred_valid)
print('10折平均Auc:{}'.format(mean_score))
lgb_pre=sum(answers)/n_folds
sub['fraud']=lgb_pre
sub.to_csv('金融2预测.csv',index=False) 

以上就是baseline的全部代码，线上提交分数是0.9463，排名显示0.95，非常简单，因为数据量也不大，训练也非常快，就做了简单的编码操作和时间戳特征，没有进行其他复杂操作，所以提升的空间还非常大，大家可以做更加复杂的特征工程，也可以深层次地研究数据业务逻辑构建有效特征，或者模型融合，这些都可以提升分数。

写在最后

本人才疏学浅，如果有写得不对或者理解错误的地方欢迎评论指正，有任何问题也可以提问，我都会一一解答！