Mobility analysis, hundreds of times faster
Fastkit-Mobility cleans GPS traces, finds stops and trips, and measures how people move. Rust does the heavy lifting; you keep your pandas or Polars dataframes.
Average time on 4 million Brightkite check-ins, Fastkit-Mobility (pandas) versus scikit-mobility. All measures and methodology.
Try it
import pandas as pd
from fastmob import radius_of_gyration
df = pd.DataFrame({
"uid": ["alice"] * 3 + ["bob"] * 3,
"datetime": pd.to_datetime(["2020-01-01 08:00", "2020-01-01 09:00", "2020-01-01 10:00"] * 2),
"lat": [41.9, 42.0, 42.1, 40.7, 40.8, 41.0],
"lng": [12.5, 12.5, 12.6, -74.0, -74.0, -73.9],
})
radius_of_gyration(df) # km each user ranges around their center
import polars as pl
from fastmob import radius_of_gyration
df = pl.DataFrame({
"uid": ["alice"] * 3 + ["bob"] * 3,
"datetime": ["2020-01-01 08:00", "2020-01-01 09:00", "2020-01-01 10:00"] * 2,
"lat": [41.9, 42.0, 42.1, 40.7, 40.8, 41.0],
"lng": [12.5, 12.5, 12.6, -74.0, -74.0, -73.9],
}).with_columns(pl.col("datetime").str.to_datetime())
radius_of_gyration(df) # returns a Polars DataFrame
From raw GPS fixes to tours
Each level of the hierarchy is derived from the one below it with a single method call.
What you can do
One day of a GeoLife user in Beijing: Fastkit-Mobility turns raw fixes into staypoints and triplegs.
Clean and segment
Filter noise, compress, and split traces into movement and stops.
Staypoints and locations
Detect where people stop and cluster recurring places such as home and work.
Measure mobility
Radius of gyration, entropy, jump lengths, flows and more, per person or per population.
Networks and map matching
Route and map-match trajectories on road and rail graphs.
Fit and evaluate
Fit mobility laws and compare generated trajectories with observed ones.
Generate synthetic mobility
Gravity, radiation, EPR and other models for flows and trajectories.
Why Fastkit-Mobility
- Familiar API. Measure names and signatures follow scikit-mobility, so existing code ports easily, and comparison tests check the results.
- Any dataframe, no copies. Narwhals accepts pandas, Polars and other eager backends and returns the same type; Rust kernels read your columns in place.
Requires Python 3.9 or later. Prebuilt wheels are available, so no Rust toolchain is needed to install; building from source is covered in development setup.
