With numerous trading models available, the system needs to identify the most appropriate model dynamically for different market conditions and product types. The key objective is to develop an automated trading system capable of selecting and adapting trading models based on real-time market data.

The accuracy and reliability of AI-based trading models heavily depend on the quality and sufficiency of the training data. The model should be able to handle unknown parameters effectively to mitigate overfitting and underfitting issues.

Incorporating unstructured data, such as news articles and blogs, into the trading model poses challenges due to the lack of standardized formats and the requirement for natural language processing.

We implemented a conjugation-based approach to create feature sets from trading information. Conjugation allows the extraction of higher order feature sets, that have proven to improve the resulting AI model.

We developed individual trading models "snipers" that analyze subsets of the available data and provide trade indications based on their specific methods. Snipers can be based on statistical, algorithmic, or AI approaches, enabling the identification of desired patterns within the data.

We enabled snipers to base their recommendations on instructed data such as news, blogs, and other relevant information sources using sentiment analysis. We also developed the snipers that focus on specific sectors, industries, or markets, allowing for targeted analysis and recommendations tailored to different trading contexts.