论文地址：
https://www.sciencedirect.com/science/article/pii/S0008884625002467
论文题目：
A transfer learning-driven paradigm for understanding cryogenic freezing mechanisms in low water/binder cement-based composites
图 1 抗弯强度预测模型结果可视化综合图：该图展示随机森林回归模型对材料抗弯强度的预测效果及数据分布特征。(a) 子图为训练集与测试集抗弯强度双向直方图，左侧红色柱形代表训练集样本数量分布，右侧深蓝色柱形代表测试集样本数量分布，直观呈现数据集的样本分布特征；(b) 子图左侧为实验值与预测值的散点拟合图，横坐标为实验抗弯强度，纵坐标为模型预测抗弯强度，黑色虚线为理想拟合线，橙色与浅橙色虚线分别代表 ±10%、±20% 误差区间，红色散点为训练集、深蓝色散点为测试集，散点越贴近理想拟合线代表模型预测精度越高，右侧为训练集与测试集预测相对误差箱线图，清晰展示两组数据的误差分布范围；(c) 子图为模型训练集与测试集的 R²、MSE、RMSE、MAE 评价指标柱状图，直观对比模型在训练与预测阶段的性能表现。正式图注：图 1 随机森林模型抗弯强度预测结果的样本分布、拟合效果及性能评价综合可视化图。
复现图片:
配色方案：
python
COLOR_SCHEMES = {
1: {'bg': '#e4eff7', 'train': '#610000', 'test': '#002060', 'line10': '#f0984d', 'line20': '#f5b881', 'r2': '#7a8e9e', 'mse': '#729fa1', 'rmse': '#a2c6d4', 'mae': '#f2aa9e'},
2: {'bg': '#f9f2ec', 'train': '#d95f02', 'test': '#7570b3', 'line10': '#1b9e77', 'line20': '#66c2a5', 'r2': '#8da0cb', 'mse': '#e78ac3', 'rmse': '#a6d854', 'mae': '#ffd92f'},
3: {'bg': '#f0f5f0', 'train': '#e41a1c', 'test': '#377eb8', 'line10': '#4daf4a', 'line20': '#98df8a', 'r2': '#9ebcda', 'mse': '#8c6bb1', 'rmse': '#88419d', 'mae': '#810f7c'},
4: {'bg': '#f7f4f9', 'train': '#984ea3', 'test': '#ff7f00', 'line10': '#a65628', 'line20': '#e41a1c', 'r2': '#b3cde3', 'mse': '#ccebc5', 'rmse': '#decbe4', 'mae': '#fed9a6'},
5: {'bg': '#f5f5f5', 'train': '#c51b7d', 'test': '#4d9221', 'line10': '#d01c8b', 'line20': '#f1b6da', 'r2': '#fbb4ae', 'mse': '#b3cde3', 'rmse': '#ccebc5', 'mae': '#decbe4'},
6: {'bg': '#edf8fb', 'train': '#8c6bb1', 'test': '#88419d', 'line10': '#8c96c6', 'line20': '#bfd3e6', 'r2': '#1b9e77', 'mse': '#d95f02', 'rmse': '#7570b3', 'mae': '#e7298a'},
7: {'bg': '#ffffcc', 'train': '#d73027', 'test': '#4575b4', 'line10': '#fc8d59', 'line20': '#fee090', 'r2': '#66c2a5', 'mse': '#fc8d62', 'rmse': '#8da0cb', 'mae': '#e78ac3'},
8: {'bg': '#f0f9e8', 'train': '#0868ac', 'test': '#43a2ca', 'line10': '#7bccc4', 'line20': '#a8ddb5', 'r2': '#a6cee3', 'mse': '#1f78b4', 'rmse': '#b2df8a', 'mae': '#33a02c'},
9: {'bg': '#fff5eb', 'train': '#7f2704', 'test': '#d94801', 'line10': '#f16913', 'line20': '#fd8d3c', 'r2': '#fb9a99', 'mse': '#e31a1c', 'rmse': '#fdbf6f', 'mae': '#ff7f00'},
10: {'bg': '#fcfbfd', 'train': '#3f007d', 'test': '#6a51a3', 'line10': '#807dba', 'line20': '#bcbddc', 'r2': '#cab2d6', 'mse': '#6a3d9a', 'rmse': '#ffff99', 'mae': '#b15928'},
11: {'bg': '#f5f0f0', 'train': '#67001f', 'test': '#053061', 'line10': '#d6604d', 'line20': '#4393c3', 'r2': '#8dd3c7', 'mse': '#ffffb3', 'rmse': '#bebada', 'mae': '#fb8072'},
12: {'bg': '#eef5e5', 'train': '#00441b', 'test': '#40004b', 'line10': '#1b7837', 'line20': '#762a83', 'r2': '#80b1d3', 'mse': '#fdb462', 'rmse': '#b3de69', 'mae': '#fccde5'},
13: {'bg': '#f7f7f7', 'train': '#252525', 'test': '#737373', 'line10': '#525252', 'line20': '#969696', 'r2': '#d9d9d9', 'mse': '#bdbdbd', 'rmse': '#969696', 'mae': '#737373'},
14: {'bg': '#f0f4f5', 'train': '#b2182b', 'test': '#2166ac', 'line10': '#d6604d', 'line20': '#92c5de', 'r2': '#f4a582', 'mse': '#fddbc7', 'rmse': '#d1e5f0', 'mae': '#4393c3'},
15: {'bg': '#faebd7', 'train': '#800000', 'test': '#000080', 'line10': '#cd5c5c', 'line20': '#f08080', 'r2': '#ffc0cb', 'mse': '#ffb6c1', 'rmse': '#ff69b4', 'mae': '#ff1493'},
16: {'bg': '#f5fffa', 'train': '#2e8b57', 'test': '#8b0000', 'line10': '#3cb371', 'line20': '#cd5c5c', 'r2': '#00fa9a', 'mse': '#00ff7f', 'rmse': '#90ee90', 'mae': '#98fb98'},
17: {'bg': '#ffebcd', 'train': '#8b4513', 'test': '#2f4f4f', 'line10': '#a0522d', 'line20': '#cd853f', 'r2': '#d2b48c', 'mse': '#f4a460', 'rmse': '#daa520', 'mae': '#b8860b'},
18: {'bg': '#f0ffff', 'train': '#008b8b', 'test': '#9932cc', 'line10': '#20b2aa', 'line20': '#ba55d3', 'r2': '#e0ffff', 'mse': '#afeeee', 'rmse': '#7fffd4', 'mae': '#40e0d0'},
19: {'bg': '#fff0f5', 'train': '#c71585', 'test': '#191970', 'line10': '#db7093', 'line20': '#ffb6c1', 'r2': '#dda0dd', 'mse': '#ee82ee', 'rmse': '#ff00ff', 'mae': '#ba55d3'},
20: {'bg': '#e6e6fa', 'train': '#4b0082', 'test': '#8b008b', 'line10': '#9370db', 'line20': '#da70d6', 'r2': '#e6e6fa', 'mse': '#d8bfd8', 'rmse': '#dda0dd', 'mae': '#ee82ee'},
21: {'bg': '#f4f6f8', 'train': '#1f77b4', 'test': '#ff7f0e', 'line10': '#2ca02c', 'line20': '#d62728', 'r2': '#9467bd', 'mse': '#8c564b', 'rmse': '#e377c2', 'mae': '#7f7f7f'},
22: {'bg': '#faf5f5', 'train': '#8dd3c7', 'test': '#ffffb3', 'line10': '#bebada', 'line20': '#fb8072', 'r2': '#80b1d3', 'mse': '#fdb462', 'rmse': '#b3de69', 'mae': '#fccde5'},
23: {'bg': '#f2faeb', 'train': '#003f5c', 'test': '#2f4b7c', 'line10': '#665191', 'line20': '#a05195', 'r2': '#d45087', 'mse': '#f95d6a', 'rmse': '#ff7c43', 'mae': '#ffa600'},
24: {'bg': '#e8f4f8', 'train': '#f44336', 'test': '#9c27b0', 'line10': '#3f51b5', 'line20': '#03a9f4', 'r2': '#009688', 'mse': '#8bc34a', 'rmse': '#ffeb3b', 'mae': '#ff9800'},
25: {'bg': '#fdfcf0', 'train': '#393b79', 'test': '#5254a3', 'line10': '#6b6ecf', 'line20': '#9c9ede', 'r2': '#637939', 'mse': '#8ca252', 'rmse': '#b5cf6b', 'mae': '#cedb9c'},
26: {'bg': '#f0e6fa', 'train': '#3182bd', 'test': '#6baed6', 'line10': '#9ecae1', 'line20': '#c6dbef', 'r2': '#e6550d', 'mse': '#fd8d3c', 'rmse': '#fdae6b', 'mae': '#fdd0a2'},
27: {'bg': '#fffaf0', 'train': '#b2df8a', 'test': '#33a02c', 'line10': '#fb9a99', 'line20': '#e31a1c', 'r2': '#fdbf6f', 'mse': '#ff7f00', 'rmse': '#cab2d6', 'mae': '#6a3d9a'},
28: {'bg': '#ebf5f0', 'train': '#1b9e77', 'test': '#d95f02', 'line10': '#7570b3', 'line20': '#e7298a', 'r2': '#66a61e', 'mse': '#e6ab02', 'rmse': '#a6761d', 'mae': '#666666'},
29: {'bg': '#f5ebec', 'train': '#7fc97f', 'test': '#beaed4', 'line10': '#fdc086', 'line20': '#ffff99', 'r2': '#386cb0', 'mse': '#f0027f', 'rmse': '#bf5b17', 'mae': '#666666'},
30: {'bg': '#fafdfa', 'train': '#4e79a7', 'test': '#f28e2b', 'line10': '#e15759', 'line20': '#76b7b2', 'r2': '#59a14f', 'mse': '#edc948', 'rmse': '#b07aa1', 'mae': '#ff9da7'},
31: {'bg': '#f5f0ea', 'train': '#9c27b0', 'test': '#673ab7', 'line10': '#3f51b5', 'line20': '#2196f3', 'r2': '#00bcd4', 'mse': '#009688', 'rmse': '#4caf50', 'mae': '#8bc34a'},
32: {'bg': '#fdf5e6', 'train': '#ffc107', 'test': '#ff9800', 'line10': '#ff5722', 'line20': '#795548', 'r2': '#9e9e9e', 'mse': '#607d8b', 'rmse': '#f44336', 'mae': '#e91e63'},
33: {'bg': '#f9f9f9', 'train': '#e41a1c', 'test': '#377eb8', 'line10': '#4daf4a', 'line20': '#984ea3', 'r2': '#ff7f00', 'mse': '#ffff33', 'rmse': '#a65628', 'mae': '#f781bf'},
34: {'bg': '#f5f7fa', 'train': '#1b9e77', 'test': '#d95f02', 'line10': '#7570b3', 'line20': '#e7298a', 'r2': '#66a61e', 'mse': '#e6ab02', 'rmse': '#a6761d', 'mae': '#666666'},
35: {'bg': '#fcfcfc', 'train': '#8dd3c7', 'test': '#bebada', 'line10': '#fb8072', 'line20': '#80b1d3', 'r2': '#fdb462', 'mse': '#b3de69', 'rmse': '#fccde5', 'mae': '#bc80bd'},
36: {'bg': '#f0f4f8', 'train': '#1f77b4', 'test': '#ff7f0e', 'line10': '#2ca02c', 'line20': '#d62728', 'r2': '#9467bd', 'mse': '#8c564b', 'rmse': '#e377c2', 'mae': '#17becf'},
37: {'bg': '#fefaf5', 'train': '#393b79', 'test': '#637939', 'line10': '#8c6d31', 'line20': '#843c39', 'r2': '#7b4173', 'mse': '#5254a3', 'rmse': '#8ca252', 'mae': '#bd9e39'},
38: {'bg': '#f3f4f6', 'train': '#e63946', 'test': '#1d3557', 'line10': '#457b9d', 'line20': '#a8dadc', 'r2': '#e76f51', 'mse': '#2a9d8f', 'rmse': '#e9c46a', 'mae': '#f4a261'},
39: {'bg': '#faf5f0', 'train': '#d53e4f', 'test': '#3288bd', 'line10': '#fdae61', 'line20': '#abdda4', 'r2': '#f46d43', 'mse': '#66c2a5', 'rmse': '#fee08b', 'mae': '#5e4fa2'},
40: {'bg': '#f0f0f5', 'train': '#8c510a', 'test': '#d8b365', 'line10': '#4d9221', 'line20': '#2166ac', 'r2': '#5ab4ac', 'mse': '#01665e', 'rmse': '#c51b7d', 'mae': '#e9a3c9'},
41: {'bg': '#fcf8f2', 'train': '#b2182b', 'test': '#ef8a62', 'line10': '#9970ab', 'line20': '#5aae61', 'r2': '#67a9cf', 'mse': '#2166ac', 'rmse': '#762a83', 'mae': '#1b7837'},
42: {'bg': '#f8f8f8', 'train': '#c51b7d', 'test': '#e9a3c9', 'line10': '#fc8d59', 'line20': '#91bfdb', 'r2': '#a1d76a', 'mse': '#4d9221', 'rmse': '#d73027', 'mae': '#4575b4'},
43: {'bg': '#f5f5f5', 'train': '#4e79a7', 'test': '#f28e2b', 'line10': '#e15759', 'line20': '#76b7b2', 'r2': '#59a14f', 'mse': '#edc948', 'rmse': '#b07aa1', 'mae': '#ff9da7'},
44: {'bg': '#fbfaf9', 'train': '#ff5e5b', 'test': '#3d348b', 'line10': '#f7b801', 'line20': '#00cecb', 'r2': '#f18701', 'mse': '#ff7a5a', 'rmse': '#76d6ff', 'mae': '#593c8f'},
45: {'bg': '#f6f7f9', 'train': '#f0a202', 'test': '#202c59', 'line10': '#d95d39', 'line20': '#2b59c3', 'r2': '#581f18', 'mse': '#d36582', 'rmse': '#253d5b', 'mae': '#8bc34a'},
46: {'bg': '#fdfdfd', 'train': '#1a535c', 'test': '#4ecdc4', 'line10': '#8a2387', 'line20': '#ff6b6b', 'r2': '#ffe66d', 'mse': '#2d3047', 'rmse': '#93b7be', 'mae': '#e07a5f'},
47: {'bg': '#f4f1f8', 'train': '#6050dc', 'test': '#d52db7', 'line10': '#ff2e93', 'line20': '#ff8b6a', 'r2': '#ffcc5c', 'mse': '#88d8b0', 'rmse': '#2f8e2b', 'mae': '#1f487e'},
48: {'bg': '#fcf4f4', 'train': '#001f3f', 'test': '#0074d9', 'line10': '#ff851b', 'line20': '#ff4136', 'r2': '#3d9970', 'mse': '#2ecc40', 'rmse': '#85144b', 'mae': '#f012be'},
49: {'bg': '#f5f6f4', 'train': '#ef476f', 'test': '#ffd166', 'line10': '#06d6a0', 'line20': '#118ab2', 'r2': '#073b4c', 'mse': '#f78c6b', 'rmse': '#83d475', 'mae': '#26547c'},
50: {'bg': '#faf5f7', 'train': '#5f0f40', 'test': '#9a031e', 'line10': '#fb8b24', 'line20': '#e36414', 'r2': '#0f4c5c', 'mse': '#5c4d7d', 'rmse': '#42b883', 'mae': '#ffbf00'},
51: {'bg': '#edf4f8', 'train': '#2b2d42', 'test': '#8d99ae', 'line10': '#19647e', 'line20': '#ef233c', 'r2': '#d90429', 'mse': '#4b3f72', 'rmse': '#ffc857', 'mae': '#119da4'},
52: {'bg': '#fbfaf0', 'train': '#006466', 'test': '#4d194d', 'line10': '#7b2cbf', 'line20': '#c1121f', 'r2': '#588157', 'mse': '#003049', 'rmse': '#d62828', 'mae': '#f77f00'},
53: {'bg': '#f2f2f2', 'train': '#b5179e', 'test': '#fca311', 'line10': '#000000', 'line20': '#14213d', 'r2': '#386641', 'mse': '#3a0ca3', 'rmse': '#ff0054', 'mae': '#8ac926'},
54: {'bg': '#fff5f5', 'train': '#0081a7', 'test': '#00afb9', 'line10': '#003049', 'line20': '#d62828', 'r2': '#f07167', 'mse': '#8d99ae', 'rmse': '#ef233c', 'mae': '#2b2d42'},
55: {'bg': '#f4f6f8', 'train': '#70d6ff', 'test': '#ff70a6', 'line10': '#ff9770', 'line20': '#ffd670', 'r2': '#3cb371', 'mse': '#8a2be2', 'rmse': '#ff1493', 'mae': '#00fa9a'},
56: {'bg': '#fcfcf0', 'train': '#54478c', 'test': '#2c699a', 'line10': '#048ba8', 'line20': '#0db39e', 'r2': '#16db93', 'mse': '#d62828', 'rmse': '#f77f00', 'mae': '#fcbf49'},
57: {'bg': '#f0faf5', 'train': '#22223b', 'test': '#8d99ae', 'line10': '#f28e2b', 'line20': '#e15759', 'r2': '#76b7b2', 'mse': '#59a14f', 'rmse': '#edc948', 'mae': '#b07aa1'},
58: {'bg': '#f9f5fa', 'train': '#ffbe0b', 'test': '#fb5607', 'line10': '#ff006e', 'line20': '#8338ec', 'r2': '#3a86ff', 'mse': '#06d6a0', 'rmse': '#118ab2', 'mae': '#073b4c'},
59: {'bg': '#f2f8fc', 'train': '#d62828', 'test': '#84a98c', 'line10': '#9b5de5', 'line20': '#00bbf9', 'r2': '#00f5d4', 'mse': '#fee440', 'rmse': '#f15bb5', 'mae': '#003049'},
60: {'bg': '#faf3f3', 'train': '#f94144', 'test': '#277da1', 'line10': '#f9c74f', 'line20': '#4d908e', 'r2': '#f3722c', 'mse': '#90be6d', 'rmse': '#43aa8b', 'mae': '#f8961e'},
}完整代码：
python
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import matplotlib.gridspec as gridspec
from matplotlib.lines import Line2D
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import r2_score, mean_squared_error, mean_absolute_error
from PIL import Image
import os
import matplotlib
# ===================== 全局字体&绘图基础配置 =====================
# 矢量图字体兼容设置
matplotlib.rcParams['pdf.fonttype'] = 42
matplotlib.rcParams['ps.fonttype'] = 42
# 全局字体：英文期刊常用 Times New Roman
plt.rcParams['font.family'] = 'Times New Roman'
# 正常显示坐标轴负号
plt.rcParams['axes.unicode_minus'] = False
# 数学公式字体样式
plt.rcParams['mathtext.fontset'] = 'stix'
# 配色方案定义
COLOR_SCHEMES = {
1: {'bg': '#e4eff7', 'train': '#610000', 'test': '#002060', 'line10': '#f0984d', 'line20': '#f5b881',
'r2': '#7a8e9e', 'mse': '#729fa1', 'rmse': '#a2c6d4', 'mae': '#f2aa9e'},
}
SCHEME_ID = 1
scheme = COLOR_SCHEMES[SCHEME_ID]
# 图片保存目录，不存在则自动创建
save_dir = "./图表1"
os.makedirs(save_dir, exist_ok=True)
# ==============================================================
# 子图1：训练集/测试集目标变量双向直方图
# ==============================================================
def plot_back_to_back_histogram(train_target, test_target):
# 设置直方图区间
bins = np.arange(4, 29, 1)
bin_centers = 0.5 * (bins[:-1] + bins[1:])
counts_train, _ = np.histogram(train_target, bins=bins)
counts_test, _ = np.histogram(test_target, bins=bins)
# 创建画布，画布尺寸(宽,高)
fig, ax = plt.subplots(figsize=(5, 6))
# 设置坐标轴背景色
ax.set_facecolor(scheme['bg'])
# 绘制左侧训练集直方图，添加label用于图例
ax.barh(bin_centers, -counts_train, height=1, color=scheme['train'],
alpha=0.7, edgecolor=scheme['train'], label='train', zorder=2)
# 绘制右侧测试集直方图，添加label用于图例
ax.barh(bin_centers, counts_test, height=1, color=scheme['test'],
alpha=0.7, edgecolor=scheme['test'], label='test', zorder=2)
# 中间分割竖线
ax.axvline(0, color='black', linewidth=1.5, zorder=3)
# X轴刻度与字体：字号16、加粗
xticks = [-30, -20, -10, 0, 10, 20, 30]
ax.set_xticks(xticks)
ax.set_xticklabels([str(abs(x)) for x in xticks], fontsize=16, fontweight='bold')
ax.set_xlim(-32, 32)
# Y轴刻度与字体：字号16、加粗
yticks = [4, 8, 12, 16, 20, 24, 28]
ax.set_yticks(yticks)
ax.set_yticklabels(yticks, fontsize=16, fontweight='bold')
ax.set_ylim(4, 28)
# 坐标轴刻度线：线宽2、长度6
ax.tick_params(axis='both', width=2, length=6)
# X轴标签：Count，字号18、加粗
ax.set_xlabel('Count', fontsize=18, fontweight='bold')
# Y轴标签：抗弯强度，字号18、加粗
ax.set_ylabel('Flexural strength (MPa)', fontsize=18, fontweight='bold')
# 子图编号(a)：字号20、加粗
ax.text(-32, 29, '(a)', fontsize=20, fontweight='bold', va='bottom', ha='left')
# 图例：位置右上，字体14、加粗，图例边框线宽1.5
legend = ax.legend(loc='upper right', prop={'weight': 'bold', 'size': 14})
legend.get_frame().set_edgecolor('black')
legend.get_frame().set_linewidth(1.5)
# 坐标轴边框：线宽2、黑色
for spine in ax.spines.values():
spine.set_linewidth(2)
spine.set_color('black')
# 仅保存PNG图片，dpi=300高清
plt.savefig(f'{save_dir}/histogram{SCHEME_ID}.png', dpi=300, bbox_inches='tight')
plt.close()
# ==============================================================
# 子图2：回归拟合散点图 + 误差箱线组合图（左右双栏子图）
# ==============================================================
def plot_prediction_and_error(y_true_train, y_pred_train, y_true_test, y_pred_test):
# 计算相对误差(百分比)
error_train = (y_pred_train - y_true_train) / y_true_train * 100
error_test = (y_pred_test - y_true_test) / y_true_test * 100
# 创建总画布，尺寸(宽,高)
fig = plt.figure(figsize=(10, 6))
# 网格布局：1行2列，宽度比例3.5:1，子图间距0.25
gs = gridspec.GridSpec(1, 2, width_ratios=[3.5, 1], wspace=0.25)
# -------- 左子图：拟合散点图 --------
ax1 = fig.add_subplot(gs[0])
ax1.set_facecolor(scheme['bg'])
# 横向网格线：虚线、浅灰色
ax1.yaxis.grid(True, linestyle='--', color='lightgray', alpha=1, zorder=0)
# 绘制训练集、测试集散点
ax1.scatter(y_true_train, y_pred_train, color=scheme['train'], s=80,
alpha=0.8, edgecolors='white', zorder=3)
ax1.scatter(y_true_test, y_pred_test, color=scheme['test'], s=80,
alpha=1, edgecolors='white', zorder=3)
# 绘制参考拟合线、±10%、±20%误差线
min_val, max_val = 2, 30
x_line = np.linspace(min_val, max_val, 100)
ax1.plot(x_line, x_line, color='black', linestyle='--', linewidth=1.8, zorder=2)
ax1.plot(x_line, x_line * 1.1, color=scheme['line10'], linestyle='--', dashes=(5, 4), linewidth=1.5, zorder=2)
ax1.plot(x_line, x_line * 0.9, color=scheme['line10'], linestyle='--', dashes=(5, 4), linewidth=1.5, zorder=2)
ax1.plot(x_line, x_line * 1.2, color=scheme['line20'], linestyle='-.', linewidth=1.5, zorder=2)
ax1.plot(x_line, x_line * 0.8, color=scheme['line20'], linestyle='-.', linewidth=1.5, zorder=2)
# 坐标轴范围 + 刻度：字号18、加粗
ax1.set_xlim(min_val, max_val)
ax1.set_ylim(min_val, max_val)
ticks_ax1 = [5, 10, 15, 20, 25, 30]
ax1.set_xticks(ticks_ax1)
ax1.set_yticks(ticks_ax1)
ax1.set_xticklabels(ticks_ax1, fontsize=18, fontweight='bold')
ax1.set_yticklabels(ticks_ax1, fontsize=18, fontweight='bold')
ax1.tick_params(width=2, length=6)
# 坐标轴标签：字号20、加粗
ax1.set_xlabel('Experimental value (MPa)', fontsize=20, fontweight='bold')
ax1.set_ylabel('Predicted value (MPa)', fontsize=20, fontweight='bold')
# 子图编号(b)：字号22、加粗
ax1.text(2, 31, '(b)', fontsize=22, fontweight='bold', va='bottom', ha='left')
# 自定义图例：字体16、加粗，无边框
legend_elements = [
Line2D([0], [0], marker='o', color='white', markerfacecolor=scheme['train'], markersize=10, label='train'),
Line2D([0], [0], marker='o', color='white', markerfacecolor=scheme['test'], markersize=10, label='test'),
Line2D([0], [0], color=scheme['line10'], linestyle='--', linewidth=2, label='10%'),
Line2D([0], [0], color=scheme['line20'], linestyle='-.', linewidth=2, label='20%')
]
ax1.legend(handles=legend_elements, loc='upper left', ncol=2, frameon=False, prop={'weight': 'bold', 'size': 16})
# 边框样式：线宽2、黑色
for spine in ax1.spines.values():
spine.set_linewidth(2)
spine.set_color('black')
# -------- 右子图：误差箱线图 --------
ax2 = fig.add_subplot(gs[1])
ax2.set_facecolor(scheme['bg'])
ax2.yaxis.grid(True, linestyle='--', color='lightgray', alpha=0.8, zorder=0)
# 绘制箱线图，填充配色
boxes = ax2.boxplot([error_test, error_train], positions=[1, 2], widths=0.4,
patch_artist=True, showfliers=False, zorder=2)
for patch, color in zip(boxes['boxes'], [scheme['test'], scheme['train']]):
patch.set_facecolor(color)
patch.set_edgecolor('white')
patch.set_linewidth(1)
for median in boxes['medians']:
median.set_color('white')
median.set_linewidth(1.5)
for part in ['whiskers', 'caps']:
for item, color in zip(boxes[part], [scheme['test']] * 2 + [scheme['train']] * 2):
item.set_color(color)
item.set_linewidth(1.5)
# 箱线图叠加散点（抖动避免重叠）
x_test_jitter = np.random.normal(1, 0.08, size=len(error_test))
x_train_jitter = np.random.normal(2, 0.08, size=len(error_train))
ax2.scatter(x_test_jitter, error_test, color=scheme['test'], s=25, alpha=0.9,
edgecolors='white', linewidths=0.5, zorder=3)
ax2.scatter(x_train_jitter, error_train, color=scheme['train'], s=25, alpha=0.9,
edgecolors='white', linewidths=0.5, zorder=3)
# 坐标轴范围 & 刻度：字号18、加粗
ax2.set_xlim(0.5, 2.5)
ax2.set_ylim(-30, 30)
ticks_ax2_y = [-30, -20, -10, 0, 10, 20, 30]
ax2.set_yticks(ticks_ax2_y)
ax2.set_yticklabels(ticks_ax2_y, fontsize=18, fontweight='bold')
ax2.set_xticks([1, 2])
ax2.set_xticklabels(['test', 'train'], fontsize=18, fontweight='bold', rotation=-30)
ax2.tick_params(width=2, length=6)
# Y轴标签：字号20、加粗
ax2.set_ylabel('Error (%)', fontsize=20, fontweight='bold', labelpad=-5)
# 边框样式：线宽2、黑色
for spine in ax2.spines.values():
spine.set_linewidth(2)
spine.set_color('black')
# 保存图片，关闭画布
plt.savefig(f'{save_dir}/prediction_and_error{SCHEME_ID}.png', dpi=300, bbox_inches='tight')
plt.close()
# ==============================================================
# 子图3：模型评价指标柱状图 + 指标名称表格（左右组合）
# ==============================================================
def plot_metrics_bar_chart(y_true_train, y_pred_train, y_true_test, y_pred_test, cv_std_errors):
# 读取训练集交叉验证指标
r2_tr = mean_r2_tr
mse_tr = mean_mse_tr
rmse_tr = mean_rmse_tr
mae_tr = mean_mae_tr
# 计算测试集指标
mae_te = mean_absolute_error(y_true_test, y_pred_test)
mse_te = mean_squared_error(y_true_test, y_pred_test)
rmse_te = np.sqrt(mse_te)
r2_te = r2_score(y_true_test, y_pred_test)
values = [r2_tr, mse_tr, rmse_tr, mae_tr, r2_te, mse_te, rmse_te, mae_te]
y_pos = [1, 2, 3, 4, 5, 6, 7, 8]
colors = [scheme['r2'], scheme['mse'], scheme['rmse'], scheme['mae']] * 2
# 创建画布，尺寸(宽,高)
fig = plt.figure(figsize=(5, 6))
# 网格布局：1行2列，宽度比例1:2.5，子图间距0
gs = gridspec.GridSpec(1, 2, width_ratios=[1, 2.5], wspace=0)
# -------- 左侧：指标名称表格 --------
ax_tab = fig.add_subplot(gs[0])
ax_tab.set_xlim(0, 1)
ax_tab.set_ylim(0.5, 8.5)
ax_tab.axis('off') # 隐藏原生坐标轴
# 绘制表格外框与分割线
ax_tab.plot([0, 1], [0.5, 0.5], color='black', lw=4)
ax_tab.plot([0, 1], [8.5, 8.5], color='black', lw=4)
ax_tab.plot([0, 0], [0.5, 8.5], color='black', lw=4)
ax_tab.plot([1, 1], [0.5, 8.5], color='black', lw=4)
ax_tab.axvline(0.35, color='black', linewidth=1.5)
for y in [1.5, 2.5, 3.5, 5.5, 6.5, 7.5]:
ax_tab.plot([0.35, 1], [y, y], color='black', linewidth=1.5)
ax_tab.axhline(4.5, color='black', linewidth=2.0)
# 文本标注：train / test，字号18、加粗、竖排
ax_tab.text(0.175, 2.5, 'train', va='center', ha='center', rotation=90, fontweight='bold', fontsize=18)
ax_tab.text(0.175, 6.5, 'test', va='center', ha='center', rotation=90, fontweight='bold', fontsize=18)
# 指标名称：R²/MSE/RMSE/MAE，字号14、加粗；r'' 原始字符串避免转义警告
labels = [r'$\mathbf{R^2}$', 'MSE', 'RMSE', 'MAE']
for i, lbl in enumerate(labels):
ax_tab.text(0.675, i + 1, lbl, va='center', ha='center', fontweight='bold', fontsize=14)
ax_tab.text(0.675, i + 5, lbl, va='center', ha='center', fontweight='bold', fontsize=14)
# -------- 右侧：指标数值柱状图 --------
ax_bar = fig.add_subplot(gs[1])
ax_bar.set_facecolor(scheme['bg'])
ax_bar.set_ylim(0.5, 8.5)
ax_bar.set_xlim(0, 4.5)
ax_bar.set_yticks([]) # 隐藏Y轴刻度
# 纵向网格线：灰色点线
ax_bar.xaxis.grid(True, linestyle=':', color='gray', alpha=0.7, zorder=0)
# 绘制横向柱状图 + 误差棒
ax_bar.barh(y_pos, values, height=0.6, color=colors, xerr=cv_std_errors,
capsize=4, error_kw={'linewidth': 1.5}, zorder=3)
# 柱子顶部标注数值，字号16、加粗
for i, val in enumerate(values):
ax_bar.text(val + cv_std_errors[i] + 0.1, y_pos[i], f"{val:.2f}",
va='center', ha='left', fontweight='bold', fontsize=16)
# X轴刻度：字号24、加粗
ticks = [0, 1, 2, 3]
ax_bar.set_xticks(ticks)
ax_bar.set_xticklabels([str(t) for t in ticks], fontweight='bold', fontsize=24)
# 边框样式：线宽1.5、黑色
for spine in ax_bar.spines.values():
spine.set_linewidth(1.5)
spine.set_color('black')
ax_bar.tick_params(axis='x', width=2, length=6)
# 子图编号(c)：字号22、加粗
fig.text(0.02, 0.915, '(c)', fontweight='bold', fontsize=22)
plt.subplots_adjust(left=0.05, right=0.95, top=0.88, bottom=0.1, wspace=0)
# 保存图片，关闭画布
plt.savefig(f'{save_dir}/bar_chart{SCHEME_ID}.png', dpi=300, bbox_inches='tight')
plt.close()
# ==============================================================
# 组合总图：将 子图1+子图2+子图3 横向拼接为一张大图
# ==============================================================
def combine_plots(scheme_id):
# 读取三张独立子图图片
path_a = f'{save_dir}/histogram{scheme_id}.png'
path_b = f'{save_dir}/prediction_and_error{scheme_id}.png'
path_c = f'{save_dir}/bar_chart{scheme_id}.png'
img_a = Image.open(path_a)
img_b = Image.open(path_b)
img_c = Image.open(path_c)
# 计算拼接后总宽度、统一高度（取三张图最大高度）
total_width = img_a.width + img_b.width + img_c.width
max_height = max(img_a.height, img_b.height, img_c.height)
# 创建空白画布，白底
combined_img = Image.new('RGB', (total_width, max_height), color=(255, 255, 255))
# 依次横向粘贴三张子图
combined_img.paste(img_a, (0, 0))
combined_img.paste(img_b, (img_a.width, 0))
combined_img.paste(img_c, (img_a.width + img_b.width, 0))
# 保存最终组合图，画质95
combined_img.save(f'{save_dir}/combined_plot_{scheme_id}.png', quality=95)
# ==============================================================
# 主程序：数据读取、模型训练、调用绘图函数
# ==============================================================
if __name__ == '__main__':
# 读取Excel数据
df_rf_data = pd.read_excel('data.xlsx', sheet_name='RandomForest_Data')
y_target = df_rf_data['Target'].values
X_features = df_rf_data.drop(columns=['Target']).values
# 划分训练集、测试集 8:2
X_train, X_test, y_train, y_test = train_test_split(X_features, y_target, test_size=0.2, random_state=42)
# 随机森林回归模型 + 网格搜索参数
rf_model = RandomForestRegressor(random_state=42)
param_grid = {'n_estimators': [50, 100], 'max_depth': [5, 10]}
scoring_metrics = {'r2': 'r2', 'mse': 'neg_mean_squared_error',
'rmse': 'neg_root_mean_squared_error', 'mae': 'neg_mean_absolute_error'}
# 5折交叉验证网格搜索
grid_search = GridSearchCV(estimator=rf_model, param_grid=param_grid, cv=5,
scoring=scoring_metrics, refit='r2', n_jobs=-1,
return_train_score=True, verbose=2)
grid_search.fit(X_train, y_train)
# 提取最优模型对应的训练集指标均值
best_index = grid_search.best_index_
mean_r2_tr = grid_search.cv_results_['mean_train_r2'][best_index]
mean_mse_tr = -grid_search.cv_results_['mean_train_mse'][best_index]
mean_rmse_tr = -grid_search.cv_results_['mean_train_rmse'][best_index]
mean_mae_tr = -grid_search.cv_results_['mean_train_mae'][best_index]
# 提取指标标准差（用于误差棒）
std_r2 = grid_search.cv_results_['std_train_r2'][best_index]
std_mse = grid_search.cv_results_['std_train_mse'][best_index]
std_rmse = grid_search.cv_results_['std_train_rmse'][best_index]
std_mae = grid_search.cv_results_['std_train_mae'][best_index]
cv_std_errors = [std_r2, std_mse, std_rmse, std_mae, 0, 0, 0, 0]
# 最优模型预测
best_model = grid_search.best_estimator_
y_pred_train = best_model.predict(X_train)
y_pred_test = best_model.predict(X_test)
# 依次绘制三张独立子图
plot_back_to_back_histogram(y_train, y_test)
plot_prediction_and_error(y_train, y_pred_train, y_test, y_pred_test)
plot_metrics_bar_chart(y_train, y_pred_train, y_test, y_pred_test, cv_std_errors)
# 拼接组合总图
combine_plots(SCHEME_ID)数据获取
评论+私信获取
VX获取：期刊复现|Python实现回归拟合散点图与误差分析图
https://mp.weixin.qq.com/s/5PtWdKhJk5ufBWABrsDd0A
参考来源：python+遥感学习日志