Abstract: Recent advances in modeling of financial markets have focused on understanding deep statistical dependencies among a large number of financial assets and their characteristics. We model joint dynamics of thousands of companies along with hundreds of their financial data fields, e.g. market prices, fundamentals and technical indicators; and when we don’t have sufficient historical data we use generative models using machine learning to produce synthetic data. We show that generative methods have a broad range of applications in finance, including generating realistic financial time-series, volatility and correlation estimation and portfolio optimization. We will demonstrate applications in Data Imputation and Now-casting particularly for deducing geographical, climate and ESG exposures of companies that fail to report on these metrics. We also apply generative modeling for general asset pricing and hedging use cases.
Bio: Arun heads the Bloomberg Quantitative Research Solutions Team in the CTO office. Arun has worked extensively on Stochastic Volatility Models for Derivatives & Exotics Valuation, as well as factor investing, asset allocation, portfolio optimization and risk models. More recently, he has enjoyed working at the intersection of quantitative finance and using AI/Machine Learning to help reveal embedded signals in traditional & alternative data such as Company Financials, ESG, News/Social, Supply Chain, and Geo-locational & Extreme Weather data among others to study their potential impact on financial markets. In an attempt to complete a full circle, Arun has most recently been exploring use of ML methods in traditional quant space, e.g. Derivatives pricing, prediction of illiquid instruments fair value, Economic indicators Now-casting & high-dimensional Data Imputation.
Prior to joining Bloomberg, he earned his PhD from Cornell University in the areas of computer science and applied mathematics and a B. Tech in Computer Science from IIT Delhi, India. Arun is also an editorial board member of The Journal of Financial Data Science.