What is it about?
digital movements was instrumental in the development of a novel and AI-supported platform to support securities traders in generating trading signals based on alternative data.
The project was successfully implemented. However, since the generated trading signals did not result in significant outperformance in the backtest, the project has not yet been published.
What did we do?
- Developed a concept for the platform in collaboration with securities trading experts
- Development of POCs for individual system components of the platform, e.g. for the preparation of data from alternative data sources (e.g. Google Trends)
- Development of a delivery and deployment pipeline for trading system models
- Integration of AI algorithms (e.g. Nearest Neighbor) into the delivery pipeline
- Development of tools for automated data retrieval and preparation
- Development of tools for visual monitoring and evaluation of trading signals
- Execution of intensive backtests and evaluation of trading signals
Which methods/tools did we use?
- Python
- ScikitLearn
- Backtrader
- PyTrends
- Pandas
- NumPy
- Google Trends
- Docker
- AWS Cloud, AWS Lambda, AWS S3
- IntelliJ PyCharm
- Jira, Confluence
What results did we achieve?
The tool was able to generate trading signals based on alternative data as specified. The backtesting results showed a positive return under the relevant market conditions, but no significant outperformance compared to the defined benchmarks.
What conclusions have we drawn?
The results show that the AI-based generation of trading signals based on alternative data, as we have implemented it, still needs to be improved. We plan to resume the project with a different approach. Among other things, we will examine whether xLSTM-based model approaches perform better.
