SR&ED Claim Information for Chen Li

Source: Notion | Last edited: 2023-11-25 | ID: 9fb0e299-b07...

---

Name: Chen Li

Position: Founder and CEO

Duration of Employment for Fiscal Year: full-year of 2023

SR&ED-able Time: 70%

Project Title: Improving Predictive Model Performance and Developing Training Data for EonPredictor

1. Project Description:

• Brief Description: This project involved developing new model structures and methodologies to enhance the performance of financial predictive models and creating sophisticated training data for the second-layer model, EonPredictor.

1. Technological Advancements and Challenges:

• Innovative Technologies/Methods: Developed a momentum-based transformer model, integrating dynamic and static features like price, volume, volatility, day of week, month of year, and symbol category. Implemented risk-adjusted returns in loss functions, shifting from traditional absolute return metrics.
• Technological Challenges/Uncertainties: Faced challenges in enhancing the contextual understanding and performance of predictive models under varying market conditions. Addressed uncertainties in optimizing exposure control during major market drawdowns.

1. Experimental Development and Analysis:

• Experimental Processes/Methods: Experimented with new model structures for predictive accuracy. Utilized a ‘rolling backfill’ method for generating training data, sequentially training models with historical data and generating PNL data for subsequent years.
• Analysis and Findings: The new methodologies led to improved model performance in terms of contextual understanding and risk-adjusted returns, and the ‘rolling backfill’ method provided a more realistic and incremental approach to training data development.

1. Results and Impact:

• Results Achieved: Enhanced predictive model performance and more accurate training data for the EonPredictor.
• Impact on Field/Industry: Contributed to the advancement of machine learning in financial modeling, particularly in the area of risk management and realistic market condition simulation.

1. Documentation and Supporting Evidence:

• Types of Documentation: Project reports, model code repositories, experimental data, and historical performance records.

1. Time Allocation:

• Estimated Time on SR&ED Activities: 100%

📍 Site Map of SR&ED Claims