Flow Forecast

Abstract

1. Introduction

Flow Forecast (FF) is an open-source deep learning for time series forecasting framework. It provides all the latest state of the art models (transformers, attention models, GRUs, ODEs) and cutting edge concepts with easy to understand interpretability metrics, cloud provider integration, and model serving capabilities. Flow Forecast was the first time series framework to feature support for transformer based models and remains the only true end-to-end deep learning for time series framework. Currently, Task-TS from CoronaWhy primarily maintains this repository. Pull requests are welcome. Historically, this repository provided open source benchmark and codes for flash flood and river flow forecasting.

For additional tutorials and examples please see our tutorials repository.

We have a number of models we are planning on releasing soon. Please check our project board for more info

2. Objective

Deep learning PyTorch library for time series forecasting, classification, and anomaly detection (originally for flood forecasting).

This project demonstrates how Python, Deep Learning, PyTorch can be applied to a real-world AI & Machine Learning problem.

4. Technology Stack

5. System Requirements

General requirements for this technology stack — check the README for exact versions.

• Python 3.8 or later
• pip / virtualenv for dependencies
• VS Code, PyCharm or Jupyter Notebook
• Git (to clone the repository)

6. Installation & Setup

git clone https://github.com/AIStream-Peelout/flow-forecast.git
cd flow-forecast

1. Run pip install flood-forecast
2. Detailed info on training models can be found on the Wiki.
3. Check out our Confluence Documentation
4. Vanilla LSTM (LSTM): A basic LSTM that is suitable for multivariate time series forecasting and transfer learning.
5. Full transformer (SimpleTransformer in model_dict): The full original transformer with all 8 encoder and decoder blocks. Requires passing the target in at inference.
6. Simple Multi-Head Attention (MultiHeadSimple): A simple multi-head attention block and linear embedding layers. Suitable for transfer learning.
7. Transformer with a linear decoder (CustomTransformerDecoder in model_dict): A transformer with n-encoder blocks (this is tunable) and a linear decoder. Suitable for forecasting, classification or anomaly detection.
8. DA-RNN: (DARNN) A well rounded model with which utilizes a LSTM + attention.

Full setup instructions are in the project README.

7. Future Enhancements

Suggested extensions you can add to make this your own project.

• Deploy the model as a web app with Streamlit, Flask or FastAPI
• Compare against an additional model and report the metric difference
• Add explainability (SHAP / Grad-CAM)

8. Viva / Review Questions

Common questions examiners ask for projects in this domain.

1. What dataset does the project use and how was it pre-processed?
2. Which algorithm / model architecture is used and why was it chosen over alternatives?
3. How are training and testing data split, and how is overfitting avoided?
4. Which evaluation metrics (accuracy, precision, recall, F1) are reported and what do they mean here?
5. How would you deploy this model for real users?

9. Source Code & License

This project is developed by AIStream-Peelout and published on GitHub under the GNU General Public License v3.0. Please follow the license terms and credit the original author when you use or modify this code.