temporian 0.1.2

Temporian is a Python package for feature engineering of temporal data, focusing on preventing common modeling errors and providing a simple and powerful API, a first-class iterative development experience, and efficient and well-tested implementations of common and not-so-common temporal data preprocessing functions.

Temporian is a Python library for feature engineering 🛠 and data augmentation ⚡ of temporal data 📈 (e.g. time-series, transactions) in machine learning applications 🤖.

Warning Temporian development is in alpha.

Key features

• Temporian operates natively on multivariate and multi-index time-series and time-sequences data. With Temporian, all temporal data processing is unified.

• Temporian favors iterative and interactive development in Colab, where users can easily visualize intermediate results 📊 each step of the way.

• Temporian introduces a novel mechanism to prevent unwanted future leakage and training/serving skew 😰. Temporian programs always return the same result in batch and in streaming mode.

• Temporian programs can run seamlessly in-process in Python, on large datasets using Apache Beam ☁️, and in streaming for continuous data ingestion.

• Temporian's core is implemented in C++ and highly optimized 🔥, so large amounts of data can be handled in-process. In some cases, Temporian can provide a speed-up in the order of 1000x compared to other libraries.

Installation

Temporian is available on PyPI. To install it, run:

pip install temporian

Minimal example

Consider the following dataset.

$ head sales.csv
timestamp,store,price,count
2022-01-01 00:00:00+00:00,CA,27.42,61.9
2022-01-01 00:00:00+00:00,TX,98.55,18.02
2022-01-02 00:00:00+00:00,CA,32.74,14.93
2022-01-02 00:00:00+00:00,TX,48.69,83.99
...

We compute the weekly sales per store as follows.

import temporian as tp

input_data = tp.from_csv("sales.csv")

# Define a Temporian program
input_node = input_data.node()
per_store = tp.set_index(input_node, "store")
weekly_sum = tp.moving_sum(per_store["price"], window_length=tp.duration.days(7))

# Execute Temporian program
output_data = weekly_sum.run({input_node: input_data})

# Plot the result
output_data.plot()

Check the Getting Started tutorial to try it out!

Documentation

The documentation 📚 is available at temporian.readthedocs.io. The 3 minutes to Temporian ⏰️ is the best way to start.

Contributing

Contributions to Temporian are welcome! Check out the contributing guide to get started.

Credits

Temporian is developed in collaboration between Google and Tryolabs.