Python中如何将X/Y/Z数据向量拟合到带X/Y分箱的二维表格

以下两种实现方案均可直接在Python环境中运行，无需依赖闭源工具，可按需调整参数适配业务场景。

方案1：逆距离加权（IDW）分配汇总（推荐，贴合逆插值需求） #

原理为每个散点的Z值会根据与相邻分箱中心的距离反比分配权重，所有散点的加权值累加后得到分箱的Z汇总结果，天然解决散点不落在分箱内的问题。

import numpy as np
import pandas as pd
from scipy.spatial import cKDTree

# 读取示例数据
df = pd.read_csv('your_data.csv', sep=';')
points = df[['X', 'Y']].values
z_values = df['Z'].values

# 自定义X、Y分箱边界，可按需调整
x_bins = np.arange(0, 11, 2)  # X分箱规则：0-2、2-4、...、8-10
y_bins = np.arange(10, 51, 10) # Y分箱规则：10-20、20-30、...、40-50

# 计算分箱中心坐标
x_centers = (x_bins[:-1] + x_bins[1:]) / 2
y_centers = (y_bins[:-1] + y_bins[1:]) / 2
grid_x, grid_y = np.meshgrid(x_centers, y_centers)
grid_points = np.vstack([grid_x.ravel(), grid_y.ravel()]).T

# 构建KD树快速查询每个散点最近的4个分箱中心
tree = cKDTree(grid_points)
distances, indices = tree.query(points, k=4)

# 计算逆距离权重，添加极小值避免距离为0报错
weights = 1 / (distances + 1e-8)
weights /= weights.sum(axis=1, keepdims=True)

# 初始化结果矩阵并加权累加Z值
result = np.zeros(grid_x.shape)
for i in range(len(points)):
    for j in range(4):
        row, col = np.unravel_index(indices[i,j], grid_x.shape)
        result[row, col] += z_values[i] * weights[i,j]

# 转换为带分箱标签的二维表格，行对应Y分箱、列对应X分箱
result_df = pd.DataFrame(
    result,
    index=[f'{y_bins[i]}-{y_bins[i+1]}' for i in range(len(y_bins)-1)],
    columns=[f'{x_bins[i]}-{x_bins[i+1]}' for i in range(len(x_bins)-1)]
)
print(result_df)

方案2：先插值再分箱聚合 #

如果需要更平滑的结果，可以先对散点做二维插值生成连续曲面，再按分箱区间汇总Z值：

import numpy as np
import pandas as pd
from scipy.interpolate import griddata

# 读取示例数据
df = pd.read_csv('your_data.csv', sep=';')
points = df[['X', 'Y']].values
z_values = df['Z'].values

# 自定义分箱边界
x_bins = np.arange(0, 11, 2)
y_bins = np.arange(10, 51, 10)

# 生成高密度插值网格
grid_x_dense, grid_y_dense = np.mgrid[0:10:100j, 10:50:100j]
# 插值方法可选'nearest'（无平滑）、'linear'（线性平滑）、'cubic'（三次平滑，效果最优）
z_dense = griddata(points, z_values, (grid_x_dense, grid_y_dense), method='cubic')

# 按分箱区间聚合插值结果
result = np.zeros((len(y_bins)-1, len(x_bins)-1))
for i in range(len(y_bins)-1):
    y_mask = (grid_y_dense >= y_bins[i]) & (grid_y_dense < y_bins[i+1])
    for j in range(len(x_bins)-1):
        x_mask = (grid_x_dense >= x_bins[j]) & (grid_x_dense < x_bins[j+1])
        cell_mask = y_mask & x_mask
        # 聚合方式可选sum、mean等，按需调整
        result[i,j] = np.nanmean(z_dense[cell_mask])

# 转换为二维表格
result_df = pd.DataFrame(
    result,
    index=[f'{y_bins[i]}-{y_bins[i+1]}' for i in range(len(y_bins)-1)],
    columns=[f'{x_bins[i]}-{x_bins[i+1]}' for i in range(len(x_bins)-1)]
)

参数调整说明 #

• 直接修改x_bins和y_bins即可自定义分箱区间、分箱数量
• IDW方案中查询的近邻数量k可调整，k越大结果平滑度越高
• 插值方案中可替换method参数适配不同平滑度需求

内容的提问来源于stack exchange，提问作者And