Research:
Agent-based Model Parameters Estimation with Wasserstein Distance with Sebastiano Manzan (Preprint) (Slides)
The novel method outperformed 4 other estimation techniques (3 likelihood-based estimation methods and the method of simulated moments) in terms of accuracy and computational cost.
Unearthing Market Dependencies: Data-Driven Neural Networks and Realized Volatility Prediction through Graph Discovery
Constructed a novel graph discovery integrated Graph Neural Network (GNN) model for predictive tasks and data mining in finance, particularly for high-dimensional data.
Model predicted short-term intraday Realized Volatility (RV) by using Limit Order Book (LOB) and trade data with a customized underlying network discovered by the data, effectively capturing spillover effects across financial assets. This model demonstrated a 15% improvement of the baseline LightGBM and Neural Net models.
Recommender System Matching Small Medium Enterprises with Influencers Using Graph Neural Networks. (Work in progress)
Constructing a multi-modal Graph Neural Network recommender system in PyTorch to match small medium enterprises and influencers to maximize marketing effectiveness.
Conference Talks:
30th Symposium of the Society for Nonlinear Dynamics & Econometrics, Orlando, March 16, 2023
28th Computing in Economics and Finance Conference , Dallas, June 19th, 2022