贝叶斯优化xgboost的超参数最简实现_xgboost回归贝叶斯调优

from hyperopt import hp
from hyperopt import hp, fmin, tpe, Trials, partial
from hyperopt.early_stop import no_progress_loss
import warnings
warnings.filterwarnings("ignore")
import numpy as np
# from OptMetrics import MyMetric
# from sklearn.ensemble import GradientBoostingClassifier
import xgboost as xgb
from sklearn.model_selection import train_test_split
from sklearn import datasets
from MyLogColor import log,LogLevel
import time
from sklearn.metrics import precision_recall_curve,auc,f1_score,roc_curve,auc

cancer=datasets.load_breast_cancer()
x=cancer.data
y=cancer.target

def Rollover(x):
    x = x.astype(bool)
    x = ~x
    x = x.astype(int)
    return x
####TODO:将少数变成正例
y = Rollover(y)

x_train, x_val, y_train, y_val = train_test_split(x, y, test_size=0.2, random_state = 42)


def ROC_AUC(test_y, proba):
    fpr,tpr,threshold = roc_curve(test_y, proba)
    roc_auc_ = auc(fpr,tpr)
    return roc_auc_


historical_metrics = []
historical_params = {}

boosters = ['gbtree','gblinear','dart']
sampling_methods = ['uniform','gradient_based']
tree_methods = ["auto","exact","approx","hist"]
refresh_leafs = [0,1]
# process_types = ["default","update"]#,"refresh","prune"]
grow_policys = ["depthwise","lossguide"]
# sample_types = ["uniform","weighted"]
normalize_types = ["tree","forest"]
rate_drops = []



param_grid_hp = {
    'boosters':hp.choice('boosters',boosters)
    ,"n_estimators":hp.quniform("n_estimators",50,1000,1)
    ,"learning_rate":hp.uniform("learning_rate",1e-5,1)
    ,"gamma":hp.quniform("gamma",0,100,1)
    ,"max_depth":hp.quniform("max_depth",6,200,1)
    ,"min_child_weight":hp.quniform("min_child_weight",0,100,1)
    ,"max_delta_step":hp.quniform("max_delta_step",0,100,1)
    ,"subsample":hp.uniform("subsample",0,1)
    # ,"sampling_method":hp.choice("sampling_method",sampling_methods)
    ,"colsample_bytree":hp.uniform("colsample_bytree",0,1)
    ,"colsample_bylevel":hp.uniform("colsample_bylevel",0,1)
    ,"colsample_bynode":hp.uniform("colsample_bynode",0,1)
    ,"lambda":hp.quniform("lambda",0,200,1)
    ,"alpha":hp.quniform("alpha",0,200,1)
    ,"tree_method":hp.choice("tree_method",tree_methods)
    # ,"scale_pos_weight":hp.uniform("scale_pos_weight",0,1000)
    ,"refresh_leaf":hp.choice("refresh_leaf",refresh_leafs)
    # ,"process_type":hp.choice("process_type",process_types)
    ,"grow_policy":hp.choice("grow_policy",grow_policys)
    ,"max_leaves":hp.quniform("max_leaves",0,10000,1)
    ,"max_bin":hp.quniform("max_bin",256,1000,1)
    ,"num_parallel_tree":hp.quniform("num_parallel_tree",1,100,1)   
}
# booster_dart_params = {
#     "sample_type":hp.choice("sample_type",sample_types)
#     ,"normalize_type":hp.choice("normalize_type",normalize_types)
#     ,"rate_drop":hp.uniform("rate_drop",0,1)
#     ,"one_drop":hp.quniform("one_drop",0,1000,1)
#     ,"skip_drop":hp.uniform("skip_drop",0,1)
# }

booster_gblinear_params = {
    
}

def PR_AUC(test_y,proba,pred):
    precision,recall,_ = precision_recall_curve(test_y,proba)
    f1 ,pr_auc = f1_score(test_y,pred),auc(recall,precision)
    return pr_auc

def hyperopt_objective(hyperopt_params): 
    params = {
        "objective":"binary:logistic"
        ,'boosters':hyperopt_params['boosters']
        ,"n_estimators":int(hyperopt_params["n_estimators"])
        ,"learning_rate":hyperopt_params["learning_rate"]
        ,"gamma":hyperopt_params["gamma"]
        ,"max_depth":int(hyperopt_params["max_depth"])
        ,"min_child_weight":int(hyperopt_params["min_child_weight"])
        ,"max_delta_step":int(hyperopt_params["max_delta_step"])
        ,"subsample":hyperopt_params["subsample"]
        ,"verbosity":0
        # ,"sampling_method":hyperopt_params["sampling_method"]
        ,"colsample_bytree":hyperopt_params["colsample_bytree"]
        ,"colsample_bylevel":hyperopt_params["colsample_bylevel"]
        ,"colsample_bynode":hyperopt_params["colsample_bynode"]
        ,"lambda":int(hyperopt_params["lambda"])
        ,"alpha":int(hyperopt_params["alpha"])
        ,"tree_method":hyperopt_params["tree_method"]
        ,"scale_pos_weight":(y_train==0).sum()/(y_train==1).sum()
        ,"refresh_leaf":hyperopt_params["refresh_leaf"]
        # ,"process_type":hyperopt_params["process_type"]
        ,"grow_policy":hyperopt_params["grow_policy"]
        ,"max_leaves":int(hyperopt_params["max_leaves"])
        ,"max_bin":int(hyperopt_params["max_bin"])
        ,"num_parallel_tree":int(hyperopt_params["num_parallel_tree"])   
    }
    # booster_dart_params = {
    #     "sample_type":hyperopt_params["sample_type"]
    #     ,"normalize_type":hp.choice("normalize_type",normalize_types)
    #     ,"rate_drop":hyperopt_params["rate_drop"]
    #     ,"one_drop":int(hyperopt_params["one_drop"])
    #     ,"skip_drop":hyperopt_params["skip_drop"]
    # }
    dtrain = xgb.DMatrix(x_train,label=y_train)
    clf = xgb.train(params=params
                   ,dtrain=dtrain
                   ,num_boost_round=100
                   ,evals=[(dtrain,"train")]
                   ,verbose_eval=False # 不显示训练信息就改False
                   # ,obj=logistic_obj
                   )
    dtest = xgb.DMatrix(x_val,label=y_val)
    xgboost_proba = clf.predict(dtest)
    # xgbosst_proba = np.nan_to_num(xgboost_proba,0)
    
    global NOW_FUC_RUN_ITER
    NOW_FUC_RUN_ITER += 1
    metric = ROC_AUC(y_val,xgboost_proba)
    historical_metrics.append(metric)
    historical_params.update({NOW_FUC_RUN_ITER-1:params})
    return - metric

def param_hyperopt(max_evals=100):

    #保存迭代过程
    trials = Trials()

    #设置提前停止
    early_stop_fn = no_progress_loss(100)

    #定义代理模型
    #algo = partial(tpe.suggest, n_startup_jobs=20, n_EI_candidates=50)
    params_best = fmin(hyperopt_objective #目标函数
                       , space = param_grid_hp #参数空间
                       , algo = tpe.suggest #代理模型
                       #, algo = algo
                       , max_evals = max_evals #允许的迭代次数
                       , verbose=True
                       , trials = trials
                       , early_stop_fn = early_stop_fn
                      )

    #打印最优参数，fmin会自动打印最佳分数
    print("\n","\n","best params: ", params_best,
          "\n")
    return params_best, trials

NOW_FUC_RUN_ITER = 0
PARAMS_BEST, Trials = param_hyperopt(600)

historical_metrics = np.array(historical_metrics)
idx = np.argmax(historical_metrics)
params = historical_params[idx]
dtrain = xgb.DMatrix(x_train,label=y_train)
clf = xgb.train(params=params
               ,dtrain=dtrain
               ,num_boost_round=100
               ,evals=[(dtrain,"train")]
               ,verbose_eval=False # 不显示训练信息就改False
               # ,obj=logistic_obj
               )
dtest = xgb.DMatrix(x_val,label=y_val)
xgboost_proba = clf.predict(dtest)
print("测试优化结果",ROC_AUC(y_val,xgboost_proba))
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
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187