Getting Started

What is NOMAD?

NOMAD (Network for Open Mobility Analysis and Data) is an open-source Python library for end-to-end processing of large-scale GPS mobility data. The library supports:

Data ingestion from various formats (CSV, Parquet, GeoJSON)
Quality control and filtering
Spatial-temporal transformation
Derivation of mobility metrics
Synthetic trajectory generation

NOMAD is part of the NSF-funded NOMAD research infrastructure project and builds on previous software resources like scikit-mobility, mobilkit, and trackintel. The goal is to provide a single, production-ready library that covers the entire processing pipeline and is suitable for analysis of massive datasets.

Key Features

Dual API: All functions implemented in Python with parallel equivalents in PySpark
Scalable: Same analysis notebook runs on a workstation or Spark cluster without API changes
Production-ready: Designed for large-scale mobility data processing
Research-focused: Aids in the replicability of existing research

Installation

Basic Installation

Install directly from GitHub:

pip install git+https://github.com/Watts-Lab/nomad.git

With Spark Support

For large-scale data processing with PySpark:

pip install git+https://github.com/Watts-Lab/nomad.git#egg=nomad[spark]

Requirements

Python 3.9 or higher
Core dependencies: pandas, geopandas, numpy, shapely, networkx
Optional: PySpark 3.4.4+ for distributed processing

Quick Start

Here’s a simple example to get started with NOMAD:

import nomad.io.base as loader

# Load mobility data
df = loader.read_data('data/sample.csv')

# Perform stop detection
from nomad.stop_detection import dbscan
stops = dbscan.detect_stops(df, eps=100, min_samples=3)

# Compute mobility metrics
from nomad.metrics import metrics
metrics_df = metrics.compute_radius_of_gyration(df)

Next Steps

Data Ingestion - Learn how to load and validate your mobility data
Filtering & Completeness - Filter and assess data completeness
Stop Detection - Detect stops and trips from GPS traces
Visit Attribution - Infer home and workplace locations
Mobility Metrics - Compute mobility metrics
Synthetic Data Generation - Generate synthetic trajectories

Examples

The examples/ folder contains Jupyter notebooks demonstrating various NOMAD features:

[1] Reading Data - Loading different data formats
[2] Filtering - Data quality control and filtering
[3] Stop Detection - Different stop detection algorithms
[4] Home Attribution - Inferring residential locations

License

For more information, visit nomad.seas.upenn.edu.