Fold

<!-- # fold --> <p align="center" style="display:flex; width:100%; align-items:center; justify-content:center;"> <a style="margin:2px" href="https://dream-faster.github.io/fold/"><img alt="Docs" src="https://img.shields.io/github/actions/workflow/status/dream-faster/fold/docs.yaml?logo=readthedocs"></a> <a style="margin:2px" href="https://codecov.io/gh/dream-faster/fold" ><img src="https://codecov.io/gh/dream-faster/fold/branch/main/graph/badge.svg?token=Z7I2XSF188"/></a> <a style="margin:2px" href="https://github.com/dream-faster/fold/actions/workflows/tests.yaml"><img alt="Tests" src="https://github.com/dream-faster/fold/actions/workflows/tests.yaml/badge.svg"/></a> <a style="margin:2px" href="https://discord.gg/EKJQgfuBpE"><img alt="Discord Community" src="https://img.shields.io/badge/Discord-%235865F2.svg?logo=discord&logoColor=white"></a> </p> <!-- PROJECT LOGO --> <br /> <div align="center"> <a href="https://dream-faster.github.io/fold/"> <img src="https://raw.githubusercontent.com/dream-faster/fold/main/docs/images/logo.svg" alt="Logo" width="90" > </a> <h3 align="center"><b>FOLD</b><br></h3> <p align="center"> Fast <b>Adaptive Time Series ML </b> Engine <br/> <b>This is an internal project - documentation is not updated anymore and substantially differ from the current API.</b> <a href="https://dream-faster.github.io/fold/"><strong>Explore the docs »</strong></a> </p> </div> <br /> <!-- INTRO -->

The<b> <a href="https://dream-faster.github.io/fold/concepts/adaptive-ml/">Adaptive ML Engine</a></b> that lets you <b>build</b>, <b>deploy and update</b> Models easily. An order of magnitude speed-up, combined with flexibility and rigour.</b> <br/>

<!-- GETTING STARTED -->

Main Features

• 10x faster Adaptive Backtesting - What does that mean?
• Composite Models made Adaptive - What does that mean?
• Distributed computing - Why is this important?
• Update deployed models (coming in May) - Why is this important?

Installation

• Prerequisites: python >= 3.8 and pip

• Install from pypi:

  pip install fold-core
  

Quickstart

You can quickly train your chosen models and get predictions by running:

from sklearn.ensemble import RandomForestRegressor
from statsforecast.models import ARIMA
from fold import ExpandingWindowSplitter, train_evaluate
from fold.composites import Ensemble
from fold.transformations import OnlyPredictions
from fold.utils.dataset import get_preprocessed_dataset

X, y = get_preprocessed_dataset(
    "weather/historical_hourly_la", target_col="temperature", shorten=1000
)

pipeline = [
    Ensemble(
        [
            RandomForestRegressor(),
            ARIMA(order=(1, 1, 0)),
        ]
    ),
    OnlyPredictions(),
]
splitter = ExpandingWindowSplitter(initial_train_window=0.2, step=0.2)
scorecard, prediction, trained_pipelines, _, _ = train_evaluate(pipeline, X, y, splitter)

(If you install krisi by running pip install krisi you get an extended report back, rather than a single metric.)

Fold is different

• Adaptive Models and Backtesting at lightning speed.<br/> <span style="color:orange;">→ fold allows to simulate and evaluate your models like they would have performed, in reality/when deployed, with clever use of paralellization and design.</span>

• Create composite models: ensembles, hybrids, stacking pipelines, easily.<br/> <span style="color:orange;">→ Underutilized, but the easiest, fastest way to increase performance of your Time Series models. </span>

• Built with Distributed Computing in mind.<br/> <span style="color:orange;">→ Deploy your research and development pipelines to a cluster with ray, and use modin to handle out-of-memory datasets (full support for modin is coming in April).</span>

• Bridging the gap between Online and Mini-Batch learning.<br/> <span style="color:orange;">→ Mix and match xgboost with ARIMA, in a single pipeline. Boost your model's accuracy by updating them on every timestamp, if desired.</span>

• Update your deployed models, easily, as new data flows in.<br/> <span style="color:orange;">→ Real world is not static. Let your models adapt, without the need to re-train from scratch.</span>

<!-- GETTING STARTED -->

Examples, Walkthroughs and Blog Posts

<table style="width:100%"> <tr> <th>Name</th> <th>Type</th> <th>Dataset Type</th> <th>Docs Link</th> <th>Colab</th> </tr> <tr> <td> ⚡️ Core Walkthrough </td> <td>Walkthrough</td> <td>Energy</td> <td> <a href='https://dream-faster.github.io/fold/walkthroughs/core_walkthrough/' target="_blank">Notebook</a> </td> <td> <a href='https://colab.research.google.com/drive/1CVhxOmbHO9PvsdHfGvR91ilJUqEnUuy8?usp=sharing' target="_blank">Colab</a> </td> </tr> <tr> <td> 🚄 Speed Comparison of Fold to other libraries </td> <td>Walkthrough</td> <td>Weather</td> <td> <a href='https://dream-faster.github.io/fold/walkthroughs/benchmarking_sktime_fold/' target="_blank"> Notebook </a> </td> <td> <a href='https://colab.research.google.com/drive/1iLXpty-j1kpDCzLM4fCsP3fLoS_DFN1C?usp=sharing' target="_blank"> Colab </a> </td> </tr> <tr> <td> 📚 Example Collection </td> <td>Example</td> <td>Weather & Synthetic</td> <td> <a href='https://dream-faster.github.io/fold/generated/gallery/' target="_blank"> Collection Link </a> </td> <td> - </td> </tr> <tr> <td> 🖋️ Why we ended up building an Adaptive ML engine for Time Series </td> <td>Blog</td> <td>Public Release Blog Post </td> <td> <a href='https://www.appliedexploration.com/p/back-to-the-future-with-time-series' target="_blank"> Blog post on Applied Exploration </a> </td> <td> - </td> </tr> </table> <br/>

Core Features

• Supports both Regression and Classification tasks.
• Online and Mini-batch learning.
• Feature selection and other transformations on an expanding/rolling window basis
• Use any scikit-learn/tabular model natively!
• Use any univariate or sequence models (wrappers provided in fold-wrappers).
• Use any Deep Learning Time Series models (wrappers provided in fold-wrappers).
• Super easy syntax!
• Probabilistic foreacasts (currently, for Classification, full support coming in April).
• Hyperparemeter optimization / Model selection. (coming in early April!)

What is Adaptive Backtesting?

It's like classical Backtesting / Time Series Cross-Validation, plus: Inside a test window, and during deployment, fold provides a way for models to update their parameters or access the last value. Learn more

Our Open-core Time Series Toolkit

Explore our Commercial License options here

Contribution

Join our <a style="margin:2px" href="https://discord.gg/EKJQgfuBpE"><img alt="Discord Community" src="https://img.shields.io/badge/Discord-%235865F2.svg?logo=discord&logoColor=white"></a> for live discussion!

Submit an issue or reach out to us on info at dream-faster.ai for any inquiries.

Licence & Usage

We want to bring much-needed transparency, speed and rigour to the process of creating Time Series ML pipelines, while also building a sustainable business, that can support the ecosystem in the long-term. Fold's licence is inbetween source-available and a traditional commercial software licence. It requires a paid licence for any commercial use, after the initial, 30 day trial period.

We also want to contribute to open research by giving free access to non-commercial, research use of fold.

Read more

Limitations

• No intermittent time series support, very limited support for missing values.
• No hierarchical time series support.