|
|
--- |
|
|
license: apache-2.0 |
|
|
--- |
|
|
|
|
|
# FinCast: A Foundation Model for Financial Time-Series Forecasting |
|
|
|
|
|
[](link-to-paper) todo |
|
|
[](LICENSE) |
|
|
[]() |
|
|
[]() |
|
|
|
|
|
This repository contains the official implementation of **FinCast**, introduced in our paper: |
|
|
|
|
|
> **FinCast: A Foundation Model for Financial Time-Series Forecasting** |
|
|
> Zhuohang Zhu, Haodong Chen, Qiang Qu, Vera Chung |
|
|
> *CIKM 2025* (Accepted) |
|
|
|
|
|
FinCast is a **decoder-only transformer** trained on over **20B financial time points** across diverse domains and temporal resolutions. |
|
|
Technical Highlights: |
|
|
- **PQ-Loss**: Joint point + probabilistic forecasting. |
|
|
- **Mixture-of-Experts (MoE)**: Specialization across domains. |
|
|
|
|
|
--- |
|
|
|
|
|
## π₯ Features |
|
|
- Foundation model for **financial time-series forecasting**, flexible input and output length. |
|
|
- Strong performance in **zero-shot**, **supervised**, and **few-shot** settings. |
|
|
- Modular architecture with **MoE** and **quantile-aware loss**. |
|
|
- Scalable to **billions of parameters** with efficient inference. |
|
|
|
|
|
--- |
|
|
|
|
|
## π¦ Installation |
|
|
|
|
|
- The model weight can be found on π€ https://huggingface.co/Vincent05R/FinCast |
|
|
- The model code can be found on https://github.com/vincent05r/FinCast-fts |
|
|
- The corresponding datasets to reproduce the results can be found on https://huggingface.co/datasets/Vincent05R/FinCast-Paper-test |
|
|
|
|
|
Run the env_setup.sh first then run the dep_install.sh. |
|
|
|
|
|
## π Experiments |
|
|
|
|
|
- run the corresponding scripts in the scripts directory to reproduce the results in the paper. The result summary can be generate using the result summary notebook in the notebook directory. |
|
|
|
|
|
|
|
|
|
|
|
## β‘ Future Updates |
|
|
|
|
|
- PEFT finetune(LORA/DORA) is done, just need to do some testing |
|
|
- Package together for easy inference |
|
|
- Covariate Inference(currently implemented the same code as timesfm) |