数据集：处理geolife数据-CSDN博客 这边的stations，找到每个station 最近的其他10个station

1 读取数据

假设已经读完了，就是locations

2 保留有用的列

locations.drop(['center','user_id'],axis=1,inplace=True)
locations

3 加载几何形状

使用 shapely 的 wkt 模块来直接从 Well-Known Text（WKT）格式的字符串中加载几何形状，这样就无需手动解析字符串了

shapely.wkt.loads 函数可以将 WKT 格式的字符串转换为 shapely 的几何对象V

locations['extent'] = locations['extent'].apply(wkt.loads)

4  保存为geodataframe

locations=gpd.GeoDataFrame(locations,geometry='extent')
locations

locations['centroid'] = locations.geometry.centroid
locations

5 创建kd树

5.1 提取centriod坐标

import numpy as np
coordinates = np.array(list(locations['centroid'].apply(lambda x: (x.x, x.y))))
coordinates

5.2 创建kd树

kdtree = KDTree(coordinates)

6 查询最近的11个（包括自己）的邻居

distances, indices = kdtree.query(coordinates, k=11)
indices
'''
array([[  0, 518, 603, ..., 526, 345, 559],
       [  1,  12,   7, ..., 204,  56,   5],
       [  2, 124, 245, ..., 133,  96,  93],
       ...,
       [904, 852, 386, ...,  57, 253, 147],
       [905, 380,  10, ..., 154, 117, 739],
       [906, 646, 880, ..., 465, 129, 219]], dtype=int64)
'''


#剔除自己
nearest_indices = indices[:, 1:]
nearest_indices
'''
array([[518, 603, 341, ..., 526, 345, 559],
       [ 12,   7,  59, ..., 204,  56,   5],
       [124, 245, 542, ..., 133,  96,  93],
       ...,
       [852, 386, 679, ...,  57, 253, 147],
       [380,  10,   3, ..., 154, 117, 739],
       [646, 880, 429, ..., 465, 129, 219]], dtype=int64)
'''

7 保存到locations里面