trackintel

trackintel is a framework for human mobility data analysis. It provides tools for data quality enhancement, multi-source data integration, quantitative analysis, mining, and visualization. The framework is organized into a processing pipeline that operates on data adhering to the trackintel data model:

• Preprocessing: Filtering, outlier detection, and imputation of missing values.
• Contextual Augmentation: Map matching and trajectory algebra-based context addition.
• Analysis: Extraction of mobility metrics, descriptors, preferences, and systematic mobility.
• Visualization and Communication: Generation of maps, charts, and other visual outputs.

Additionally, the framework includes packages for user mobility profiling and trajectory-based learning analytics. Note that non-standardized methods are explicitly marked as experimental.

The examples/data directory contains several datasets used to demonstrate different tracking sources and formats:

• gpsies_trajectory.csv: Manually generated via GPSies.
• posmo_trajectory.csv: Automatically tracked using POSMO Segments.
• geolife_trajectory.csv and pfs_tutorial.geojson: From the Microsoft Geolife trajectory dataset.
• google_trajectory.csv: A random excerpt from a Google Maps location track.

trackintel provides 5 core classes that subclass pandas.DataFrame or geopandas.GeoDataFrame. This means you can use all standard GeoPandas operations on trackintel objects, while also benefiting from built-in data validation and specialized mobility analysis methods.

To begin processing data, you can load a CSV into a Positionfixes object, which then grants access to generation methods like generate_staypoints().

import trackintel

# Load raw data into a Positionfixes object
df = trackintel.read_positionfixes_csv('data.csv')

# Access trackintel-specific methods
df.generate_staypoints()

The trackintel framework organizes tracking data into a hierarchy of classes that represent different levels of mobility abstraction. Data is typically generated sequentially starting from raw GPS points:

1. positionfixes: The base layer containing raw GPS data (at minimum: user_id, tracked_at, longitude, latitude).
2. staypoints: Locations where a user remained for a minimum duration.
3. triplegs: Segments of movement covered using a single mode of transport.
4. locations: Meaningful places formed by clustering staypoints (can be generated by one or multiple users).
5. trips: Segments of movement between two consecutive activity staypoints (staypoints where activity=True).
6. tours: Sequences of trips that start and end at the same location. If the journey column is True, this location is considered home.

The trackintel.analysis module provides several tools for identifying significant locations in mobility data:

• location_identifier: The primary function for identifying locations.
• pre_filter_locations: Used to clean or filter location data before identification.
• freq_method: A method for location identification based on frequency.
• osna_method: A method for location identification (likely referring to a specific algorithmic approach).

You can run the provided examples directly using Python. These examples demonstrate common use cases such as preprocessing trajectories, importing/exporting data to PostGIS, and setting up example databases. Some examples, like preprocess_trajectories.py, will automatically generate output plots in the examples/out directory.

If an example requires a database connection, you can customize the connection settings by modifying the config.json file.

python preprocess_trajectories.py
python import_export_postgis.py
python setup_example_database.py

Install the trackintel package using pip. After installation, you can verify the installation by importing the library and calling print_version().

pip install trackintel

import trackintel as ti
ti.print_version()

You can install trackintel using either conda-forge (recommended) or pip. If using pip, ensure you have GeoPandas installed in your environment first.

# Recommended: Install via conda-forge
conda install -c conda-forge trackintel

# Alternative: Install via pip
pip install trackintel

To set up the trackintel data model in a PostGIS database, you can use the provided SQL commands or run the initialization script available in the repository. This ensures your database schema is compatible with the framework's data model requirements.

https://github.com/mie-lab/trackintel/blob/master/sql/create_tables_pg.sql

For data persistence, it is recommended to use a PostGIS database.

1. Execute the provided SQL script (sql/create_tables_pg.sql) in a new or empty PostGIS database to create the necessary tables.
2. Use a PostgreSQL connection string to interact with the database via trackintel functions.

trackintel includes predefined readers for common datasets to simplify the ingestion process.

Geolife

Supports easy parsing of the Geolife dataset, including the ability to add available mode labels.

• read_geolife: Parses the Geolife dataset.
• geolife_add_modes_to_triplegs: Adds mode labels to triplegs derived from Geolife data.

GPX

Used to load multiple tracks belonging to the same user.

• read_gpx: Loads GPX track data.

import trackintel

# Load Geolife data
geolife_data = trackintel.read_geolife(path_to_geolife)

# Load GPX tracks
gpx_data = trackintel.read_gpx(path_to_gpx)