You need to enable JavaScript to run this app. 基于Pandas DataFrame实现3D离散数据选区等尺度立方体可视化
优惠活动
大模型
产品
解决方案
定价
更多

基于Pandas DataFrame实现3D离散数据选区等尺度立方体可视化

3D等比例空间可视化实现方案

核心变形问题解决逻辑

  • 直接使用原始数值作为3D轴坐标会因为三个轴的量级差(X范围0.8、Y范围99、Z范围9.99),触发matplotlib自动调整轴比例,最终呈现变形的立方体
  • 解决方式:对三个轴分别做线性归一化,将全量取值范围映射到[0,1]的统一视觉区间,同时通过set_box_aspect([1,1,1])强制设置3D坐标系的空间比例为1:1:1,从根源上消除视觉变形
  • 轴刻度保留原始实际数值,不影响正常读数
    注意:如果不手动设置轴比例参数,即使完成坐标归一化,matplotlib仍会根据画布尺寸自适应调整轴长度,无法实现真正的等尺度视觉效果。

边界定义

根据给定的离散值集合和筛选条件,直接推导实际边界值,无需动态计算:

  • 外层全量空间边界:
    • X轴:[0.1, 0.9]
    • Y轴:[1, 100]
    • Z轴:[0.01, 10.0]
  • 内层目标筛选区域边界:
    • X轴:满足X>0.2 & X<=0.7的离散值范围为[0.3, 0.7]
    • Y轴:满足Y>24 & Y<=80的离散值范围为[25, 75]
    • Z轴:满足Z>0.1 & Z<=1.0的离散值为单值1.0,无实际厚度

可直接运行的代码实现

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

# 定义离散取值集合
X_vals = [0.1, 0.3, 0.5, 0.7, 0.9]
Y_vals = [1, 25, 50, 75, 100]
Z_vals = [0.01, 0.1, 1.0, 10.0]

# 模拟1000行随机分布的全量DataFrame(实际使用时替换为你自己的df即可)
np.random.seed(42)
df = pd.DataFrame({
    'X': np.random.choice(X_vals, size=1000),
    'Y': np.random.choice(Y_vals, size=1000),
    'Z': np.random.choice(Z_vals, size=1000)
})

# 定义各轴的全量实际范围
axis_ranges = {
    'X': (min(X_vals), max(X_vals)),
    'Y': (min(Y_vals), max(Y_vals)),
    'Z': (min(Z_vals), max(Z_vals))
}

# 定义实际值转视觉坐标的映射函数
def to_visual(val, axis):
    vmin, vmax = axis_ranges[axis]
    return (val - vmin) / (vmax - vmin)

# 辅助函数:绘制指定边界的半透明立方体
def draw_cube(ax, x0, x1, y0, y1, z0, z1, color, alpha):
    # 转换为视觉坐标
    vx0, vx1 = to_visual(x0, 'X'), to_visual(x1, 'X')
    vy0, vy1 = to_visual(y0, 'Y'), to_visual(y1, 'Y')
    vz0, vz1 = to_visual(z0, 'Z'), to_visual(z1, 'Z')
    # 绘制棱边
    edges = [
        ((vx0,vy0,vz0),(vx1,vy0,vz0)), ((vx1,vy0,vz0),(vx1,vy1,vz0)),
        ((vx1,vy1,vz0),(vx0,vy1,vz0)), ((vx0,vy1,vz0),(vx0,vy0,vz0)),
        ((vx0,vy0,vz1),(vx1,vy0,vz1)), ((vx1,vy0,vz1),(vx1,vy1,vz1)),
        ((vx1,vy1,vz1),(vx0,vy1,vz1)), ((vx0,vy1,vz1),(vx0,vy0,vz1)),
        ((vx0,vy0,vz0),(vx0,vy0,vz1)), ((vx1,vy0,vz0),(vx1,vy0,vz1)),
        ((vx1,vy1,vz0),(vx1,vy1,vz1)), ((vx0,vy1,vz0),(vx0,vy1,vz1))
    ]
    for p1, p2 in edges:
        ax.plot3D(*zip(p1, p2), color=color, linewidth=1.2)
    # 绘制半透明表面
    xx, yy = np.meshgrid([vx0, vx1], [vy0, vy1])
    ax.plot_surface(xx, yy, np.full_like(xx, vz0), color=color, alpha=alpha)
    ax.plot_surface(xx, yy, np.full_like(xx, vz1), color=color, alpha=alpha)
    yy, zz = np.meshgrid([vy0, vy1], [vz0, vz1])
    ax.plot_surface(np.full_like(yy, vx0), yy, zz, color=color, alpha=alpha)
    ax.plot_surface(np.full_like(yy, vx1), yy, zz, color=color, alpha=alpha)
    xx, zz = np.meshgrid([vx0, vx1], [vz0, vz1])
    ax.plot_surface(xx, np.full_like(xx, vy0), zz, color=color, alpha=alpha)
    ax.plot_surface(xx, np.full_like(xx, vy1), zz, color=color, alpha=alpha)

# 初始化画布
fig = plt.figure(figsize=(8,8))
ax = fig.add_subplot(projection='3d')
# 强制等比例设置,保障视觉尺度一致
ax.set_box_aspect([1,1,1])
ax.set_xlim(0,1)
ax.set_ylim(0,1)
ax.set_zlim(0,1)

# 绘制外层半透明全量空间立方体
draw_cube(ax,
          x0=axis_ranges['X'][0], x1=axis_ranges['X'][1],
          y0=axis_ranges['Y'][0], y1=axis_ranges['Y'][1],
          z0=axis_ranges['Z'][0], z1=axis_ranges['Z'][1],
          color='#888888', alpha=0.15
)

# 计算筛选区域边界(补极小Z方向厚度避免无厚度平面)
target_bounds = {
    'X': (0.3, 0.7),
    'Y': (25, 75),
    'Z': (1.0, 1.0)
}
z_pad = 0.05 * (axis_ranges['Z'][1] - axis_ranges['Z'][0])
# 绘制内层深色目标区域立方体
draw_cube(ax,
          x0=target_bounds['X'][0], x1=target_bounds['X'][1],
          y0=target_bounds['Y'][0], y1=target_bounds['Y'][1],
          z0=target_bounds['Z'][0] - z_pad, z1=target_bounds['Z'][1] + z_pad,
          color='#222222', alpha=0.6
)

# 叠加样本散点
df_visual = pd.DataFrame({
    'x_v': df['X'].apply(lambda v: to_visual(v, 'X')),
    'y_v': df['Y'].apply(lambda v: to_visual(v, 'Y')),
    'z_v': df['Z'].apply(lambda v: to_visual(v, 'Z')),
    'in_target': (df['X']>0.2) & (df['X']<=0.7) & (df['Y']>24) & (df['Y']<=80) & (df['Z']>0.1) & (df['Z']<=1.0)
})
# 非目标样本
ax.scatter(df_visual[~df_visual['in_target']]['x_v'],
           df_visual[~df_visual['in_target']]['y_v'],
           df_visual[~df_visual['in_target']]['z_v'],
           c='#cccccc', s=8, alpha=0.5)
# 目标样本
ax.scatter(df_visual[df_visual['in_target']]['x_v'],
           df_visual[df_visual['in_target']]['y_v'],
           df_visual[df_visual['in_target']]['z_v'],
           c='#ff3333', s=20, alpha=0.9)

# 设置轴刻度,替换为实际数值
ax.set_xticks([to_visual(v, 'X') for v in X_vals])
ax.set_xticklabels([str(v) for v in X_vals])
ax.set_yticks([to_visual(v, 'Y') for v in Y_vals])
ax.set_yticklabels([str(v) for v in Y_vals])
ax.set_zticks([to_visual(v, 'Z') for v in Z_vals])
ax.set_zticklabels([str(v) for v in Z_vals])
ax.set_xlabel('X')
ax.set_ylabel('Y')
ax.set_zlabel('Z')

plt.tight_layout()
plt.show()

展示效果说明

  • 外层为浅灰色半透明立方体,覆盖全量数据空间
  • 内层为深灰色半透明立方体,标注筛选得到的目标区域
  • 叠加的散点直观展示全量样本分布,其中红色点为被筛选逻辑命中的样本
  • 三个轴的视觉长度完全一致,不存在数值量级差异带来的变形问题

内容的提问来源于stack exchange,提问作者Ali Kazmi

相关产品推荐
方舟 Agent Plan

超全模态模型 × Harness 升级,最新支持 Deepseek-V4.1-Flash、GLM-5.3 系列、Doubao-Seedream-5.0-pro、Kimi-K3 (部分), 限时 9.9 元起

最近更新时间:2026.11.06 17:45:00