XGBoost

训练并保存模型

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

import pandas as pd
import xgboost as xgb
from sklearn.model_selection import train_test_split

# Prepare training data and setting parameters
iris = pd.read_csv('/home/ma-user/work/iris.csv')
X = iris.drop(['variety'],axis=1)
y = iris[['variety']]
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1234565)
params = {
    'booster': 'gbtree',
    'objective': 'multi:softmax',
    'num_class': 3,
    'gamma': 0.1,
    'max_depth': 6,
    'lambda': 2,
    'subsample': 0.7,
    'colsample_bytree': 0.7,
    'min_child_weight': 3,
    'silent': 1,
    'eta': 0.1,
    'seed': 1000,
    'nthread': 4,
}
plst = params.items()
dtrain = xgb.DMatrix(X_train, y_train)
num_rounds = 500
model = xgb.train(plst, dtrain, num_rounds)
model.save_model('/tmp/xgboost.m')

训练前请先下载iris.csv数据集，解压后上传至Notebook本地路径/home/ma-user/work/。iris.csv数据集下载地址：https://gist.github.com/netj/8836201。Notebook上传文件操作请参见上传本地文件至Notebook中。

保存完模型后，需要上传到OBS目录才能发布。发布时需要带上config.json配置和推理代码customize_service.py。config.json编写请参考模型配置文件编写说明，推理代码请参考推理代码。

推理代码

在模型代码推理文件customize_service.py中，需要添加一个子类，该子类继承对应模型类型的父类，各模型类型的父类名称和导入语句如请参考表1。

# coding:utf-8
import collections
import json
import xgboost as xgb
from model_service.python_model_service import XgSklServingBaseService
class UserService(XgSklServingBaseService):

    # request data preprocess
    def _preprocess(self, data):
        list_data = []
        json_data = json.loads(data, object_pairs_hook=collections.OrderedDict)
        for element in json_data["data"]["req_data"]:
            array = []
            for each in element:
                array.append(element[each])
            list_data.append(array)
        return list_data

    #   predict
    def _inference(self, data):
        xg_model = xgb.Booster(model_file=self.model_path)
        pre_data = xgb.DMatrix(data)
        pre_result = xg_model.predict(pre_data)
        pre_result = pre_result.tolist()
        return pre_result

    # predict result process
    def _postprocess(self,data):
        resp_data = []
        for element in data:
            resp_data.append({"predictresult": element})
        return resp_data