贝叶斯优化逻辑回归的代码实现

import pandas as pd
import numpy as np
# import lightgbm as lgb
from sklearn.model_selection import train_test_split
from sklearn import metrics
# from log_color import log,LogLevel
from sklearn import metrics
import time
import datetime
from hyperopt import hp
from sklearn.model_selection import KFold, cross_validate
from hyperopt import hp, fmin, tpe, Trials, partial
from hyperopt.early_stop import no_progress_loss
from sklearn.metrics import classification_report
from sklearn.linear_model import LogisticRegression
from matplotlib import pyplot as plt



titanic_table = pd.read_csv('titanic_features_table.csv')
X = titanic_table[['年龄', '兄弟姐妹/配偶的个数', '父母/小孩个数', '票价', '1等舱', '2等舱', '3等舱', '登船港口为空',
       '登船为S港', '登船为Q港', '登船为nan港', '女性', '男性',]]
y = titanic_table['是否生还']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=65)

bast_params = {
    
}

param_grid_hp = {
    'penalty':hp.choice('penalty',['l1','l2','elasticnet','none']),
    'dual':hp.choice('dual',[True,False]),
    'fit_intercept':hp.choice('fit_intercept', [True,False]),
    'solver-0':hp.choice('solver-0',['newton-cg', 'lbfgs', 'sag', 'saga']),
    'solver-1':hp.choice('solver-1',['liblinear', 'saga']),
    'solver-2':hp.choice('solver-2',['saga']),
    'tol': hp.uniform('tol',0,1),
    'C':hp.uniform('C',0,1),
    'intercept_scaling':hp.uniform('intercept_scaling',0,1),
    'max_iter':hp.uniformint('max_iter',100,1000),
    'l1_ratio':hp.uniform('il1_ratio',0,1),
    'random_state':hp.randint('random_state',100)
}

def hyperopt_objective(hyperopt_params):
    func_start = time.time()
    
    params = {
        
    }
    if hyperopt_params['penalty'] =='l2' or hyperopt_params['penalty'] == 'none':
        solver_params = hyperopt_params['solver-0']
    elif hyperopt_params['penalty'] == 'l1':
        solver_params = hyperopt_params['solver-1']
    elif hyperopt_params['penalty'] == 'elasticnet':
        solver_params = hyperopt_params['solver-2']
    
    if solver_params == 'liblinear' and hyperopt_params['fit_intercept'] ==True:
        intercept_scaling_params = hyperopt_params['intercept_scaling']
    else:
        intercept_scaling_params = 1
    
    if solver_params == 'liblinear' and hyperopt_params['penalty'] == 'l2':
        dual_params = True
    else:
        dual_params = False
        
    if hyperopt_params['penalty'] == 'elasticnet':
        l1_ratio_params = hyperopt_params['l1_ratio']
    else:
        l1_ratio_params = None
    try:
        clf = LogisticRegression(
            penalty = hyperopt_params['penalty'],
            dual = dual_params,
            fit_intercept = hyperopt_params['fit_intercept'],
            solver = solver_params,
            tol=hyperopt_params['tol'],
            C = hyperopt_params['C'],
            intercept_scaling = intercept_scaling_params,
            max_iter = hyperopt_params['max_iter'],
            l1_ratio = l1_ratio_params,
            random_state=hyperopt_params['random_state']
            )
        clf.fit(X_train, y_train)
        predict_y = clf.predict(X_test)
        acc = (predict_y==y_test).sum()/y_test.size
        bast_params.update(
            {acc:{
            'penalty' :hyperopt_params['penalty'],
            'dual': dual_params,
            'fit_intercept':hyperopt_params['fit_intercept'],
            'solver':solver_params,
            'tol':hyperopt_params['tol'],
            'C':hyperopt_params['C'],
            'intercept_scaling':intercept_scaling_params,
            'max_iter':hyperopt_params['max_iter'],
            'l1_ratio':l1_ratio_params,
            'random_state':hyperopt_params['random_state']
        }})
    except Exception as e:
        
        acc = 0
    # print(f"acc:{acc}")
    func_end = time.time()
    return acc
    
    
def param_hyperopt(max_evals=100):

    #保存迭代过程
    trials = Trials()

    #设置提前停止
    ## 如果损失没有增加，将在 X 次迭代后停止的停止函数
    early_stop_fn = no_progress_loss(400)

    #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
NUM_EVALS = 800
params_best, trials = param_hyperopt(NUM_EVALS)

print(f"所有搜索相关记录：{trials.trials[0]}")
params_best

print(f"最好的参数是：{bast_params[max(bast_params)]}\n分数是：{max(bast_params)}")