ESCP Chine Association

AI & ML interests

Research Focus Our team is currently focused on time series data modeling and deep learning optimization for cryptocurrency market forecasting. Our research integrates Granger causality analysis and multitask neural networks (with N-BEATS as the core architecture), while also incorporating a framework inspired by Fc-GAGA (Granger-Gated Attention Graph Aggregator). This approach allows us to capture temporal causal relationships between variables while leveraging graph-based attention mechanisms to dynamically adjust the model’s focus on key variables, thereby enhancing its ability to model complex time series. In addition, we have implemented efficient feature extraction methods, including STFT and custom kernel U-Net, to further capture both frequency-domain and time-domain characteristics of the data. Our goal is to achieve high-precision market forecasting with lower computational costs, while also exploring its potential applications in real-time trading systems and risk management. Our team includes: A PhD in Bayesian methods, specializing in probabilistic reasoning and uncertainty analysis; A master's student from Dauphine, specializing in financial market economics; A graduate researcher from the Dauphine Institute of Applied Mathematics, focusing on time series and high-dimensional data modeling; Myself, an ESCP student with over three years of experience in quantitative strategy research. I am responsible for leading the research direction, designing key methodologies, and coordinating interdisciplinary team efforts to ensure the success of our projects.(www.linkedin.com/in/kai-wang-a811ab1ba) Dr. Pablo Winant is also currently very interested in. Upcoming Event In February 2025, we plan to organize a one-month offline data research event, leveraging the resources of ESCP Blue Factory and the ESCP Chine association. This event will focus on the following themes: Developing and testing efficient frameworks inspired by Granger causality and attention-based mechanisms (e.g., Fc-GAGA) to enhance model adaptability and precision in complex financial systems; Applications of nonlinear extensions of Granger causality in financial data forecasting; Optimization and implementation of multitask deep learning models (such as Kernel U-Net and N-BEATS). We aim for this activity not only to advance our team’s research progress but also to attract more researchers and students interested in related fields to participate. Sincere Invitation Given your esteemed background and expertise, we sincerely invite you to follow our research and provide guidance or advice on our direction and activities. If convenient, we would be delighted if you could participate in the event or share your academic insights with us.

Team members 1