python偏最小化二乘回归（PLSR）代码附示例数据集

import numpy as np from sklearn.cross_decomposition import PLSRegression from sklearn.metrics import mean_squared_error, mean_absolute_error, r2_score import pandas as pd from sklearn.preprocessing import StandardScaler def adjusted_r_squared(r_squared, num_observations, num_predictors): return 1 - (1 - r_squared) * (num_observations - 1) / (num_observations - num_predictors - 1) df = pd.read_excel('C:/Users/12441/Desktop/大论文/反射率.xlsx', sheet_name='5-1') # 文件路径和工作表名称 # print(df.head) # 查看前5行数据 X = df.iloc[:, :6].values # 读取前6列数据 # print(X) Y = df['体积'] # 列名称，指定某一列数据 x = np.array(X) y = np.array(Y) # 数据中心化(centering)操作 # scaler = StandardScaler() # x = scaler.fit_transform(x) # y = y - y.mean() # 实例化一个 PLSRegression 对象 pls = PLSRegression(n_components=6, scale=False, copy=True) # scale，是否进行数据标准化（缩放）操作，默认为True # 训练模型 pls.fit(x, y) # 预测结果 y_pred = pls.predict(x) # 也可以单独预测另一个数据集 # 计算误差 mse = mean_squared_error(y, y_pred) mae = mean_absolute_error(y, y_pred) rmse = np.sqrt(mean_squared_error(y, y_pred)) r2 = r2_score(y, y_pred) # f_value = (r2 / (1 - r2)) * (n - n_components - 1) / n_components # n是样本数 f_value = (r2 / (1 - r2)) * (17 - 6 - 1) / 6 r_squared = r2 num_observations = 17 num_predictors = 6 adj_r_squared = adjusted_r_squared(r_squared, num_observations, num_predictors) intercept = pls.y_mean - np.dot(pls.x_mean, pls.coef_) # 打印评估指标 print(f"MSE: {mse:.4f}") # 保留4位小数 print(f"MAE: {mae:.4f}") print(f"RMSE: {rmse:.4f}") print(f"R2: {r2:.4f}") print(f"Adjusted R2: {adj_r_squared:.4f}") print(f"F: {f_value:.4f}") print(pls.coef_) print("截距：", intercept)