8

I have a dataframe of X,Y coordinates that represent points along the paths taken by several different entities. Pseudo-data here, but it is roughly of the form:

entity_id   lat   lon   time

1001        34.5  14.2  4:55 pm
1001        34.7  14.5  4:58 pm
1001        35.0  14.6  5.03 pm

1002        27.1  19.2  2:01 pm
1002        27.4  19.3  2:08 pm
1002        27.4  19.9  2:09 pm

What I would like to do is group these points by entity_id, and then arrange the points sequentially in time to create a LineString object for each entity_id. The output will be several lines/paths, with each corresponding to an entity_id.

I can do this by looping through each entity_id and each point in entity_id and using the instructions provided here, but is there a faster/more efficient way to do this leveraging GeoPandas or Shapely, perhaps with groupby?

15

I think I found an interim solution, which I'm posting in case it's useful for anyone:

import pandas as pd
import numpy as np
from geopandas import GeoDataFrame
from shapely.geometry import Point, LineString

# Zip the coordinates into a point object and convert to a GeoDataFrame
geometry = [Point(xy) for xy in zip(df.lon, df.lat)]
df = GeoDataFrame(df, geometry=geometry)

# Aggregate these points with the GroupBy
df = df.groupby(['entity_id'])['geometry'].apply(lambda x: LineString(x.tolist()))
df = GeoDataFrame(df, geometry='geometry')

Note that if you have single-point trajectories in your data, you will have to discard these first or LineString will throw an error.

This and this post were helpful in writing the GroupBy function.


Update: If you didn't discard the single point, you can also use the conditional sentence like:

 df = df.groupby(['entity_id'])['geometry'].apply(lambda x: LineString(x.tolist()) if x.size > 1 else x.tolist())
  • that is amazing! – Ufos Aug 30 '18 at 17:04
  • In case you only want the grouped dataset and to keep ID as a column, this should help: df.groupby('entity_id', as_index=False).agg({'geometry': lambda x: ...}) – Ufos Aug 30 '18 at 17:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.