机器学习算法，可识别停机时间

Question

我收到了车辆停车的实时地理坐标，并且我想在数据中标识位置热点以定义地理围栏。我已经尝试过使用DBSCAN了：

from sklearn.cluster import DBSCAN
from sklearn import metrics

# represent GPS points as (lat, lon)
coords = df_out.as_matrix(columns=['glat', 'glng'])
# eps is the max distance that points can be from each other to be considered in a cluster
# min_samples is the minimum cluster size (everything else is classified as noise)
db = DBSCAN(eps=0.01/6371, min_samples=35, algorithm='ball_tree', metric='haversine').fit(np.radians(coords))

但是它无法正确群集，并且每次我上传具有500k +行的数据时都会冻结