python fit函数_Gaussian NB fit（）函数需要固定长度的向量

import numpy as np import pandas as pd import matplotlib.pyplot as plt import seaborn as sns from sklearn.gaussian_process import GaussianProcessRegressor from sklearn.gaussian_process.kernels import RBF, Matern, ConstantKernel from skopt import gp_minimize from skopt.space import Real from skopt.utils import use_named_args from skopt.plots import plot_convergence, plot_objective from scipy.optimize import minimize from sklearn.preprocessing import StandardScaler import warnings from sklearn.exceptions import ConvergenceWarning # 设置美观的绘图风格 sns.set(style="whitegrid", palette="muted", font_scale=1.1) plt.rcParams['font.family'] = 'DejaVu Sans' plt.rcParams['axes.labelsize'] = 12 plt.rcParams['xtick.labelsize'] = 10 plt.rcParams['ytick.labelsize'] = 10 # 读取数据 df = pd.read_excel('FEM_data.xlsx') print("数据预览:") print(df.head()) # 检查数据分布 plt.figure(figsize=(12, 8)) sns.pairplot(df, diag_kind='kde') plt.suptitle('Feature and Target Distribution', y=1.02) plt.tight_layout() plt.savefig('data_distribution.png', dpi=300) plt.show() # 分离特征和目标 X = df[['f_nb', 'E', 'po', 'v_r', 'e_0.2', 'P', 'r']].values S = df['S'].values.reshape(-1, 1) # 需要最小化 RM = df['RM'].values.reshape(-1, 1) # 需要最大化 # 标准化特征和目标 scaler_X = StandardScaler().fit(X) scaler_S = StandardScaler().fit(S) scaler_RM = StandardScaler().fit(RM) X_scaled = scaler_X.transform(X) S_scaled = scaler_S.transform(S).ravel() RM_scaled = scaler_RM.transform(RM).ravel() # 定义组合目标：S越小越好，RM越大越好 -> 最小化 [S_scaled - RM_scaled] combined_target = S_scaled - RM_scaled # 为贝叶斯优化定义参数空间 param_space = [ Real(0.1, 0.9, name='f_nb'), # Continuous variable Real(120000, 280000, name='E'), # Young's modulus Real(0.27, 0.34, name='po'), # Poisson's ratio Real(0.314, 10.314, name='v_r'), # Volume ratio Real(680, 875, name='e_0.2'), # Yield strength Real(0, 4000, name='P'), # Load Real(21, 25, name='r') # Radius ] # 创建GPR模型 - 进一步扩大核参数边界防止收敛问题 def create_kernel(): return ConstantKernel(1.0, constant_value_bounds=(1e-6, 1e3)) * Matern( length_scale=1.0, length_scale_bounds=(1e-6, 1e3), nu=2.5 ) + ConstantKernel(1.0, constant_value_bounds=(1e-6, 1e3)) * RBF( length_scale=0.5, length_scale_bounds=(1e-6, 1e3) ) kernel = create_kernel() gpr = GaussianProcessRegressor( kernel=kernel, alpha=1e-5, # 减小噪声项 n_restarts_optimizer=25, # 增加重启次数 random_state=42, normalize_y=True # 内部标准化目标值 ) # 训练初始GPR模型 gpr.fit(X_scaled, combined_target) print(f"初始GPR模型参数: {gpr.kernel_}") # 定义贝叶斯优化目标函数 @use_named_args(param_space) def objective(**params): # 提取参数值 param_values = np.array([params[dim.name] for dim in param_space]).reshape(1, -1) # 标准化输入 param_scaled = scaler_X.transform(param_values) # 预测组合目标 mean_pred, std_pred = gpr.predict(param_scaled, return_std=True) # 获取置信下界 (Lower Confidence Bound) - 平衡探索与开发 lcb = mean_pred - 0.5 * std_pred return lcb[0] # 返回标量值 # 运行贝叶斯优化 - 使用8个初始点（2的幂） res_gp = gp_minimize( func=objective, dimensions=param_space, n_calls=50, # 额外评估次数 n_random_starts=8, # 使用2的幂（8个）初始点 initial_point_generator='sobol', # 更好的空间覆盖 random_state=42, acq_func='EI' ) print("\n贝叶斯优化结果:") print(f"最优参数组合: {res_gp.x}") print(f"最佳目标值: {res_gp.fun}") # 获取最优参数 optimal_params = np.array(res_gp.x).reshape(1, -1) optimal_params_scaled = scaler_X.transform(optimal_params) # 预测最优点的组合目标值 combined_pred = gpr.predict(optimal_params_scaled) # 分别预测S和RM的标准化值 # 创建新的GPR模型用于S和RM gpr_S = GaussianProcessRegressor( kernel=create_kernel(), # 使用相同的核结构 alpha=1e-5, n_restarts_optimizer=25, # 增加重启次数 random_state=42, normalize_y=True ) gpr_S.fit(X_scaled, S_scaled) gpr_RM = GaussianProcessRegressor( kernel=create_kernel(), # 使用相同的核结构 alpha=1e-5, n_restarts_optimizer=25, # 增加重启次数 random_state=42, normalize_y=True ) gpr_RM.fit(X_scaled, RM_scaled) # 预测最优点的S和RM S_pred_scaled = gpr_S.predict(optimal_params_scaled) RM_pred_scaled = gpr_RM.predict(optimal_params_scaled) # 反标准化得到实际值 S_pred = scaler_S.inverse_transform(S_pred_scaled.reshape(-1, 1)) RM_pred = scaler_RM.inverse_transform(RM_pred_scaled.reshape(-1, 1)) print("\n预测最优值:") print(f"S_min: {S_pred[0][0]:.2f}") print(f"RM_max: {RM_pred[0][0]:.2f}") # 可视化优化过程 plt.figure(figsize=(12, 8)) plot_convergence(res_gp) plt.title('Bayesian Optimization Convergence', fontsize=14) plt.tight_layout() plt.savefig('bo_convergence.png', dpi=300) plt.show() # 绘制目标函数响应面 - 传递所有维度名称 dimension_names = [dim.name for dim in param_space] plt.figure(figsize=(16, 12)) ax = plot_objective(res_gp, dimensions=dimension_names) plt.suptitle('Objective Function Response Surface', fontsize=16) plt.tight_layout(rect=[0, 0, 1, 0.96]) # 为标题留出空间 plt.savefig('objective_response.png', dpi=300) plt.show() # 分析参数敏感性 sensitivity = [] # 创建包装函数，处理参数传递问题 def wrapped_objective(params): """包装函数，处理参数传递问题""" param_dict = {dim.name: val for dim, val in zip(param_space, params)} return objective(**param_dict) # 使用中心差分法计算梯度 delta = 0.05 base_params = np.array(res_gp.x) for idx, dim in enumerate(param_space): # 上限参数值 up_params = base_params.copy() up_params[idx] = min(up_params[idx] + delta * (dim.bounds[1] - dim.bounds[0]), dim.bounds[1]) up_score = -wrapped_objective(up_params) # 下限参数值 low_params = base_params.copy() low_params[idx] = max(low_params[idx] - delta * (dim.bounds[1] - dim.bounds[0]), dim.bounds[0]) low_score = -wrapped_objective(low_params) sensitivity.append({ 'parameter': dim.name, 'absolute_sensitivity': up_score - low_score, 'relative_sensitivity': (up_score - low_score) / (abs(up_params[idx] - low_params[idx]) + 1e-10), 'range': dim.bounds }) # 创建参数敏感性数据框 sensitivity_df = pd.DataFrame(sensitivity).sort_values('absolute_sensitivity', ascending=False) # 绘制参数敏感性 plt.figure(figsize=(12, 7)) sns.barplot( y='parameter', x='absolute_sensitivity', data=sensitivity_df, palette='viridis' ) plt.xlabel('Sensitivity to Objective Function', fontsize=12) plt.ylabel('Parameter', fontsize=12) plt.title('Parameter Sensitivity Analysis', fontsize=14) plt.tight_layout() plt.savefig('parameter_sensitivity.png', dpi=300) plt.show() # 打印完整优化报告 print("\n=================== OPTIMIZATION REPORT ===================") print(f"最优参数组合 (原始值):") for dim, val in zip(param_space, res_gp.x): print(f"- {dim.name}: {val:.4f} (范围: [{dim.bounds[0]}, {dim.bounds[1]}])") # 计算实际提升百分比 S_mean = df['S'].mean() RM_mean = df['RM'].mean() S_improvement = (S_mean - S_pred[0][0]) / S_mean * 100 RM_improvement = (RM_pred[0][0] - RM_mean) / RM_mean * 100 print(f"\n目标提升预测:") print(f"S最小值 (预测): {S_pred[0][0]:.2f} (降幅: {S_improvement:.1f}%)") print(f"RM最大值 (预测): {RM_pred[0][0]:.2f} (增幅: {RM_improvement:.1f}%)") print("\n参数敏感性排序:") print(sensitivity_df[['parameter', 'absolute_sensitivity']].to_string(index=False)) # 保存优化结果 params_names = [dim.name for dim in param_space] results_df = pd.DataFrame([res_gp.x], columns=params_names) results_df['S_pred'] = S_pred[0][0] results_df['RM_pred'] = RM_pred[0][0] results_df.to_csv('optimization_results.csv', index=False) print("\n优化结果已保存到 optimization_results.csv")报错：TypeError: objective() got an unexpected keyword argument 'f_nb'

python利用curve_fit进行拟合