This is a software that handles the next scenarios
- Download and preprocess historical data for stock markets
- Create Machine learning model for prediction future price movement
- Tune hyperparameters of the model and data preprocessing
- Use UI to manage training/tuning jobs
- Save results of every execution for future analysis
- Pick best runs and deploy for productive trading
All together this solution covers end-to-end process.
So far there is no pip package published so the only way now is to clone this repo
In order to launch optimization jobs you would need to
- download
orderlog
files for needed tickers from here - convert it from
qsh
tobin
format via this tool
Either local filesystem (fast, not scalable) or S3 storage
Create a Mongo instance and to add credentials to ./py/config.py
. Some hosting options with free tiers
./py
folder has python scripts with main logic./flask_ui
scripts that do start a server with UI for scheduling experiments and tracking their results./docker
has Dockerfiles for lua tests (bot) and for running tuning jobs
The easiest way is to add experiments via UI on locally hosted Flask server and locally launch a Docker container that will start worker. And worker will pick the next job that is ready for processing
- TA-lib
- TA list of indicators
- Stockstats python module
- Backtrader
- Backtesting Systematic Trading Strategies in Python: Considerations and Open Source Frameworks
- Habr finam
- Benchmarking
- Reddit post
- https://scikit-optimize.github.io
- Scikit optimize examples
- BOHB: ROBUST AND EFFICIENT HYPERPARAMETER OPTIMIZATION AT SCALE
- HpBandSter
- От «ЦЕРИХ» (Plaza II)
- ftp от Zerich
- ftp://athistory.zerich.com/
- От «ФИНАМ» (Plaza II)
- От «Scalping.Pro»
- Архив данных
- Format conversion
- Trades from Quik via ODBC
- Py quik
- Pandas from MySQL
- oAuth service
- tweak mysql odbc
- Открытие обучение
- TP and SL explained
- Lua Quik RPC Python