Cryptocurrency forecasting 📈 training and serving models made automatic
Project description
🚧 Be carefull this repo is still a work in progress
What is already functional?
- Prefect Flows - 80%
- Training pipeline - 100%
- Serving models - 100%
- Interface - 100%
- CLI - 70%
- Documentation - 0%
Dev package available on PyPI.
Make Us Rich
Deep Learning applied to cryptocurrency forecasting.
For more details on how to use this project, please refer to documentation.
You can inspect the training pipeline with the Kedro Viz
tool, available here
Introduction
We provide a simple way to train, serve and use cryptocurrency forecasting models on a daily basis.
Every hour Prefect
flows are launched to train and store models automatically.
Each flow has 2 variables: currency
and compare
to identify which type of data the fetching data
part
needs to get from the Binance API
to train the model.
For example, if you want to train a model on the currency Bitcoin
compared with US Dollar
: currency="btc",compare="usdt"
.
You have to give the symbol for each variable. Find all available symbols on the Binance platform.
Once the Kedro
pipeline is launched, everything works smoothly and automatically.
There is 5 steps for the pipeline to complete:
- 🪙 Fetching data from Binance API.
- 🔨 Preprocessing data:
- Extract features from fetched data.
- Split extracted features.
- Scale splitted features.
- Create sequences with scaled train features.
- Create sequences with scaled test features.
- Split train sequences as train and validation sequences.
- 🏋️ Training model.
- 🔄 Converting model to ONNX format.
- 📁 Uploading converted model to object storage service.
After the end of the training pipeline, the new model will be available for serving.
Every 10 minutes, a Prefect
flow is launched to update the API with lastest available models for each currency.
The final step is the crypto dashboard that allows users to see forecasting for their favorite assets.
Prerequisites
The main project has poetry
as package manager. If you need to install poetry, check their awesome
documentation.
You don't need to clone this project to your local machine. You can simply install the make-us-rich
package with this
command:
$ pip install make-us-rich
It's recommended to have an isolated environment for each component of the project, unless you run everything on the same machine.
CLI Usage
TODO
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for make_us_rich-0.1.0.dev5-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | bf675eb13efccb81243160c7015db29200bbe617d950bec86f6b121ec0e3b533 |
|
MD5 | 8d892c93ea17a385a235f738811d5995 |
|
BLAKE2b-256 | d30a2d01f67611809847a11920cce7e0f94bebed05e66019811b6f54e2239aec |