详见github
https://github.com/chengsong990020186/xgboost_score_model/tree/master
如对你有学习有帮助，请支持点赞~
主要内容：
- 1.使用xgboost训练模型，并保存。
- 2.基于falsk框架，生成实时api接口,进行部署。
- 3.测试api接口。
- 4.api文档
训练数据已上传至github，可以自行下载。
import pandas as pd
import xgboost as xgb
from sklearn.model_selection import train_test_split
from sklearn.externals import joblib
import logging
data=pd.read_csv('C:\\Users\\HP\\Desktop\\give me some credit\\data\\cs-training.csv')
del data['Unnamed: 0']
data.columns=['y','RevolvingUtilizationOfUnsecuredLines', 'age','NumberOfTime30-59DaysPastDueNotWorse', 'DebtRatio', 'MonthlyIncome','NumberOfOpenCreditLinesAndLoans', 'NumberOfTimes90DaysLate','NumberRealEstateLoansOrLines', 'NumberOfTime60-89DaysPastDueNotWorse','NumberOfDependents']
data_x=data[['RevolvingUtilizationOfUnsecuredLines', 'age','NumberOfTime30-59DaysPastDueNotWorse', 'DebtRatio', 'MonthlyIncome','NumberOfOpenCreditLinesAndLoans', 'NumberOfTimes90DaysLate','NumberRealEstateLoansOrLines', 'NumberOfTime60-89DaysPastDueNotWorse','NumberOfDependents']]
data_y=data['y']
train_x, test_x, train_y, test_y = train_test_split(data_x.values, data_y.values, test_size=0.2,random_state=1234)
d_train = xgb.DMatrix(train_x, label=train_y)
d_valid = xgb.DMatrix(test_x, label=test_y)
watchlist = [(d_train, 'train'), (d_valid, 'valid')]
#参数设置
params={
'eta': 0.2, # 特征权重 取值范围0~1 通常最后设置eta为0.01~0.2
'max_depth':3, # 通常取值：3-10 树的深度
'min_child_weight':1, # 最小样本的权重，调大参数可以防止过拟合
'gamma':0.3,
'subsample':0.8, #随机取样比例
'colsample_bytree':0.8, #默认为1 ，取值0~1 对特征随机采集比例
'booster':'gbtree', #迭代树
'objective': 'binary:logistic', #逻辑回归，输出为概率
# 'nthread':12, #设置最大的进程量，若不设置则会使用全部资源
'scale_pos_weight': 1, #默认为0，1可以处理类别不平衡
'lambda':1, #默认为1
'seed':1234, #随机数种子
'silent':1 , #0表示输出结果
'eval_metric': 'auc' # 检验指标
}
bst = xgb.train(params, d_train,1000,watchlist,early_stopping_rounds=500, verbose_eval=10)
tree_nums=bst.best_ntree_limit
logging.info('最优模型树的数量：%s，auc：%s' % (bst.best_ntree_limit, bst.best_score))
bst = xgb.train(params, d_train,tree_nums,watchlist,early_stopping_rounds=500, verbose_eval=10)
joblib.dump(bst, 'd:/xgboost.model') #保存模型
在目录建立 api.py 文件，代码如下
# -*- coding: utf-8 -*-
from flask import Flask, jsonify,request
import xgboost as xgb
from sklearn.externals import joblib
import numpy as np
import logging
import logging.handlers
logger = logging.getLogger()
file_handler = logging.handlers.TimedRotatingFileHandler('api_20190530.log', when='midnight')
# 测试按秒保留3个旧log文件
# file_handler = logging.handlers.TimedRotatingFileHandler("test.log", when='S', interval=1, backupCount=3)
# 设置后缀名称，跟strftime的格式一样
# file_handler.suffix = "%Y-%m-%d_%H-%M-%S.log"
formatter = logging.Formatter('%(asctime)s - %(message)s')
file_handler.setFormatter(formatter)
logger.addHandler(file_handler)
logger.setLevel(logging.INFO)
#模型使用的特征
col_x=['RevolvingUtilizationOfUnsecuredLines',
'age',
'NumberOfTime30-59DaysPastDueNotWorse',
'DebtRatio',
'MonthlyIncome',
'NumberOfOpenCreditLinesAndLoans',
'NumberOfTimes90DaysLate',
'NumberRealEstateLoansOrLines',
'NumberOfTime60-89DaysPastDueNotWorse',
'NumberOfDependents']
xgboost_score_model=joblib.load('d:/xgboost.model')
app = Flask(__name__)
app.config["JSONIFY_PRETTYPRINT_REGULAR"] = False
app.config['JSON_AS_ASCII'] = False
@app.route('/xgboost_score_model/v1.0', methods=['POST'])
def create_task():
logger.info(str(request.get_data()))
try:
request_data=request.get_json() #获取传入数据
user_data=[request_data['user_data'][col] for col in col_x]
result={}
proba=xgboost_score_model.predict(xgb.DMatrix(np.array(user_data).reshape(1,-1)))[0] #模型预测概率
B=20/(np.log(0.5))
A=600-B*(np.log(1/9))
score=A+B*np.log(proba/(1-proba)) #概率转化为评分
result['code']=1
result['user_id']=request_data['user_id']
result['proba']=str(proba)
result['score']=str(score)
logger.info(result)
return jsonify(result) #返回结果
except KeyError as e:
logger.info(e)
return jsonify({'code':101,'error_type':'KeyError','error_msg':'输入特征错误:'+str(e)})
except ValueError as e:
logger.info(e)
return jsonify({'code':102,'ValueError':'ValueError','error_msg':'输入值错误：'+str(e)})
except Exception as e:
logger.info(e)
return jsonify({'code':103,'error_type':'Unchek_Exception','error_msg':'未知错误:'+str(e)})
if __name__ == '__main__':
# app.run(debug=True)
app.run(host='0.0.0.0', port=8000, debug=True)
在终端运行代码,进入Python文件目录，运行：
python ./api.py
api接口开启成功，可以开始实时调用，并保存调用日志。
运行以下代码进行测试
import requests
import json
data_sample='''
{
"user_id": "123456789",
"user_data": {"DebtRatio": 0.80,
"MonthlyIncome": 9120.0,
"NumberOfDependents": 2.0,
"NumberOfOpenCreditLinesAndLoans": 13.0,
"NumberOfTime30-59DaysPastDueNotWorse": 2.0,
"NumberOfTime60-89DaysPastDueNotWorse": 0.0,
"NumberOfTimes90DaysLate": 0.0,
"NumberRealEstateLoansOrLines": 6.0,
"RevolvingUtilizationOfUnsecuredLines": 0.766,
"age": 45.0}
}
'''
data=json.loads(data_sample)
url='http://127.0.0.1:8000/xgboost_score_model/v1.0'
page=requests.post(url,json=data)
print(page.json())
该接口是评分模型调用。
测试环境：http://127.0.0.1:8000/xgboost_score_model/v1.0
正式环境：***
POST
JSON
| 参数名称 | 参数类型 | 描述 |
|---|---|---|
| user_id | string | 用户id |
| user_data | json | 用户特征数据 |
{
"user_id": "123456789",
"user_data": {"DebtRatio": 0.8029821290000001,
"MonthlyIncome": 9120.0,
"NumberOfDependents": 2.0,
"NumberOfOpenCreditLinesAndLoans": 13.0,
"NumberOfTime30-59DaysPastDueNotWorse": 2.0,
"NumberOfTime60-89DaysPastDueNotWorse": 0.0,
"NumberOfTimes90DaysLate": 0.0,
"NumberRealEstateLoansOrLines": 6.0,
"RevolvingUtilizationOfUnsecuredLines": 0.76612660900000007,
"age": 45.0}
}
格式 : JSON 编码 : UTF-8 返回数据示例:
{
"code": 1,
"proba": "0.397148",
"score": "548.644005001",
"user_id": "123456789"
}
{
"code": 101,
"error_msg": "输入特征错误:'aaaa'",
"error_type": "KeyError"
}
{
"ValueError": "ValueError",
"code": 102,
"error_msg": "输入值错误：invalid literal for int() with base 10: ''"
}
{
"code": 103,
"error_msg": "未知错误:400 Bad Request: Failed to decode JSON object: Expecting ',' delimiter: line 12 column 1 (char 227)",
"error_type": "Unchek_Exception"
}
| 字段名 | 字段名称 | 描述 |
|---|---|---|
| code | 状态码 | 1：返回成功 |
| user_id | 用户id | 用户id |
| proba | 预测概率 | 预测概率 |
| score | 预测分数 | 预测分数 |
评分模型的开发、部署、测试、文档说明就基本完成了。如有疑问可与我联系：[email protected]