sifeng 0.5.0

The machine learning quant research framework.

Si Feng: The machine learning quant research framework

Documentation: Getting Started | API Reference
Releases: PyPI | Source | Changelog

Sifeng (思风, imported as sifeng) is a quant research framework built by our team, which also share the same name Sifeng. It is a python package mainly designed to be used within our team but also welcomes public users. Feel free to tell us what features you want to add to this project. Currently it supports many features such as:

• :book: Detailed wiki, and annotations.
• :chart_with_upwards_trend: Accessible and easy to use function for retrieving structurized data.
• :zap: SOTA deep learning models, bayesian nerual networks, useful tools, and so on.

Table of Contents

• Table of Contents
• Installation
• Highlights

Installation

It is recommended that you install using pip, and make sure to install the latest possible version to aviod any problems:

$ pip install sifeng            # normal install
$ pip install --upgrade sifeng  # upgrade if needed

Alternatively, it is also possible to install sifeng by cloning this repository:

$ git clone https://github.com/zty200489/sifeng.git
$ cd sifeng
$ pip install .

sifeng requires the following dependencies:

• Numpy
• Pandas
• Duckdb
• joblib
• requests
• tqdm
• pytorch
• colorama

Highlights

• :chart_with_upwards_trend: We understand that in quant resarch, data is the key. So we provide easy-to-use API for retrieving data from remote OSS server. We currently only supports daily HLOC data, indicators and basic information, but we will include other kinds of data shortly in the future.
• :handshake: As out deep-learning models are based on torch, we support uniform and consistent APIs that can be pieced easily into your eexisting projects to ease you the burden of repetitive hard work.
• :bookmark_tabs: Detailed and fully-customizeable logs that fits specifically for your project. You can easily design your own log format or print custom log messages by inheriting the VerboseBase class and passing it into an trainer.
• :abacus: Model is the second most important part when doing quant researches. So we provide SOTA deep-learning models including MultiheadedSelfAttentionModule, MixtureOfExpertsBlock and even Bayesian modules to help you conduct research across various kinds of data.