TanaT 0.8.0

A library for temporal analysis of trajectories.

TanaT

Temporal Analysis of Trajectories

TanaT is a powerful Python library designed for advanced temporal sequence analysis, with specialized focus on patient care pathways and complex temporal data structures (trajectories).

Stay Updated

Subscribe to our newsletter to get updates, release notes, and example notebooks straight to your inbox!

What Makes TanaT Different

TanaT bridges the gap between traditional time series analysis and complex temporal sequence modeling by offering:

• Expressive Data Representation: Handle event sequences, interval sequences, and state sequences with unified APIs
• Advanced Distance Metrics: Specialized metrics for temporal data including DTW, edit distance, and custom metrics
• Flexible Clustering: State-of-the-art clustering algorithms adapted for temporal sequences and trajectories
• Extensible Architecture: Modular design allowing easy integration of new methods and metrics

Core Capabilities

Data Structures

• Event Sequences: Point-in-time events with rich feature descriptions
• Interval Sequences: Time-spanning events with overlapping support
• State Sequences: Continuous state representations with temporal transitions
• Trajectories: Multi-dimensional temporal data combining multiple sequence types

Analysis Methods

• Distance Computation: Dynamic Time Warping, Edit Distance, Longest Common Subsequence, and more
• Clustering: Specialized algorithms for grouping similar temporal patterns
• Filtering & Selection: Advanced criteria-based data selection and manipulation
• Visualization: Comprehensive tools for temporal data exploration
• Survival analysis: Model and predict time until key events

Scientific Foundation

TanaT draws inspiration from established frameworks:

• TraMineR (R): State sequence analysis methodologies
• aeon & tslearn: Time series analysis best practices

Architecture Overview

TanaT provides a comprehensive suite of interconnected modules for end-to-end temporal sequence analysis:

Feature	Description
Simulation	Generate synthetic data for statistical power analysis and algorithm benchmarking
Visualization	Explore and interpret temporal sequences through rich visual representations
Data Wrangling	Manipulate, filter, and transform temporal data with flexible operations
Survival Analysis	Integrate time-to-event modeling and survival techniques
Metrics & Clustering	Apply specialized distance metrics and clustering algorithms for temporal data
Workflow Orchestration	Build reproducible, automated analysis pipelines

Resources

• Documentation: Full Documentation
• Source Code: GitLab Repository
• Issues & Support: Issue Tracker

Citation

If you use TanaT in your research, please cite:

@inproceedings{tanat2025,
title={Towards a Library for the Analysis of Temporal Sequences},
authors={Thomas Guyet and Arnaud Duvermy},
booktitle={Proceedings of AALTD, ECML Workshop on Advanced Analytics and Learning on Temporal Data},
year={2025},
pages={16}
}

Affiliation & Support

TanaT is actively developed within the AIstroSight Inria Team.

The development has been supported by:

• 2024-2025: AIRacles Chair (Inria/APHP/CS)
• 2025-present: PEPR/SafePaw project (Government funding managed by the French National Research Agency under France 2030, reference number ANR-22-PESN-0005)

Team

Core Development Team

• Arnaud Duvermy - Architecture & Core Development
• Thomas Guyet - Project Leadership & Research Methods

Contact: TanaT

This work benefits from the advice of Mike Rye.

---

TanaT is open source software designed to advance temporal sequence analysis in research and industry applications.