Corporate Climate assessment: an applied NLP-clustering approach

Abstract:

The interest of financial practitioners and policymakers for climate risk assessment of corporates is increased drastically during the last years. In the first part of this talk, we describe different scoring approaches for the assessment of climate performance of corporates and the ‘translation’ of such assessments into adjusted portfolios. In the second part, we focus on one particular dimension of climate performance, i.e., the quality of disclosures. We develop an NLP driven approach for the automatic extraction of the climate-related topics and for the clustering of companies publications based on their climate content.

Bio:

Giuseppe Bonavolonta’ works as senior validator for the EIB Group, where he is involved in the assessment and validation of the market, credit, ALM, liquidity, capital, counterparty and climate risk models. He has a long standing experience and previously worked in financial institutions as quantitative analyst and risk manager.

Giuseppe Bonavolonta’ earned a PhD in Mathematics from the university of Luxembourg with a summa cum laude distinction.

Oleg Reichmann works as senior model developer at the European Central Bank currently focusing on climate risk modelling for corporates. Previously, he worked as senior validator for the EIB Group, model developer for the Swiss National Bank as well as postdoctoral researcher and lecturer at ETH Zurich.

Oleg Reichmann holds a PhD in Applied Mathematics from ETH Zurich, Switzerland as well as a master in Mathematics and one in Business Management from University of Muenster, Germany.

View Slides

A Word is Worth a Thousand Dollars: Adversarial Attack on Tweets Fools Stock Prediction

Abstract:

More and more investors and machine learning models rely on social media (e.g., Twitter and Reddit) to gather real-time information and sentiment to predict stock price movements. Although text-based models are known to be vulnerable to adversarial attacks, whether stock prediction models have similar vulnerability is underexplored. In this talk, I report on our NAACL’22 paper’s result with an experiment with a variety of adversarial attack configurations to fool three stock prediction victim models. In this experiment, we address the task of adversarial generation by solving combinatorial optimization problems with semantics and budget constraints. Our results show that the proposed attack method can achieve consistent success rates and cause significant monetary loss in trading simulation by simply concatenating a perturbed but semantically similar tweet.

Bio:

Dakuo Wang is a Senior Research Staff Member at IBM Research, Principal Investigator at MIT-IBM Watson AI Lab, and Adjunct Professor at Northeastern University. His research lies at the intersection of human-computer interaction (HCI), artificial intelligence (AI), and computer-supported team collaboration (CSCW), with a focus on the exploration, development, and evaluation of human-centered AI (HCAI) systems within the real-world context. The overarching research goal is to democratize AI for every person and every organization, so that they can easily access AI and collaborate with AI to accomplish real-world tasks better -- the “human-AI collaboration” paradigm. Before joining IBM Research, Dakuo got his Ph.D. from the University of California Irvine (“how people write together now” co-advised by Judith Olson and Gary Olson). He has worked as a designer, researcher, and engineer in the U.S., China, and France. He has served in various organizing committees, program committees, and editorial boards for conferences and journals, and ACM has recognized him as a Distinguished Speaker.

View Slides

Iterated and exponentially weighted moving principal component analysis

Abstract:

The principal component analysis (PCA) is a staple statistical and unsupervised machine learning technique in finance. The application of PCA in a financial setting is associated with several technical difficulties, such as numerical instability and nonstationarity. We attempt to resolve them by proposing two new variants of PCA: an iterated principal component analysis (IPCA) and an exponentially weighted moving principal component analysis (EWMPCA). Both variants rely on the Ogita–Aishima iteration as a crucial step.

Bio:

Dr Paul Bilokon is CEO and Founder of Thalesians Ltd and an academic at Imperial College, where his work focuses on machine learning, high performance computing, big data, and electronic trading. His career in quantitative finance spans Morgan Stanley, Lehman Brothers, Nomura, Citigroup, Deutsche Bank, and BNP Paribas, and he and his team at Thalesians continue to provide consulting services to numerous financial institutions, both on the buy-side and on the sell-side. He was one of the e-credit pioneers and has co-authored several books, including Machine Learning in Finance: From Theory to Practice and Big Data and High-Frequency Data with kdb+/q. Paul is fluent in C++, Java, Python, and kdb+/q and enjoys building distributed software systems powered by ML and applied mathematics.

View Slides

Machine Learning in Exotic Derivatives Trading and Hedging

Abstract:

The past 5 years have witnessed a boom in the structured products issuances on US stocks, driven by a low rates environment and a rallying market, making the yield enhancement that these products allow very attractive. The scale of this business, coupled with the complexity of the risk and pricing models and the need for speed, have emphasized the demand for machine learning as an alternative way for pricing and managing complex equities risk. The purpose of this presentation is to provide an overview of the structured products business on US stocks and to highlight the areas in which machine learning can advance the pricing and risk management as well as the benefits traders expect from it.
 

Bio:

Dr. Amal Moussa is a Managing Director at Goldman Sachs where she leads the Single Stocks Exotic Derivatives Trading desk. Prior to that, Amal held senior level positions in equity derivatives trading at other leading financial institutions such as J.P. Morgan, UBS and Citigroup. In addition to her work in Markets, Amal is an Adjunct Professor at Columbia University where she teaches a graduate course on Modeling and Trading Derivatives in the Mathematics of Finance Masters program.

Amal has a Ph.D. in Statistics, obtained with distinction, from Columbia University. Her thesis “Contagion and Systemic Risk in Financial Networks” shed light on the importance of the network structure in identifying systemic financial institutions and formulating regulatory policies, and has been cited by several scholars and industry professionals including former Federal Reserve president Janet Yellen. She was also awarded the Minghui Yu Teaching Award at Columbia University. Prior to her Ph.D., Amal graduated with a Masters in Mathematical Finance from Sorbonne University (former Paris VI) and a Grande Ecole engineering degree from Télécom Paris. Amal is a board member of Teach for Lebanon, an NGO working to ensure that all children in Lebanon have access to education regardless of socioeconomic background, and she is an active member of the Women in Trading network at Goldman Sachs.
 

View Slides

Deep Learning Statistical Arbitrage

Abstract:

Statistical arbitrage identifies and exploits temporal price differences between similar assets. We propose a unifying conceptual framework for statistical arbitrage and develop a novel deep learning solution, which finds commonality and time-series patterns from large panels in a data-driven and flexible way. First, we construct arbitrage portfolios of similar assets as residual portfolios from conditional latent asset pricing factors. Second, we extract the time series signals of these residual portfolios with one of the most powerful machine learning time-series solutions, a convolutional transformer. Last, we use these signals to form an optimal trading policy, that maximizes risk-adjusted returns under constraints. We conduct a comprehensive empirical comparison study with daily large cap U.S. stocks. Our optimal trading strategy obtains a consistently high out-of-sample Sharpe ratio and substantially outperforms all benchmark approaches. It is orthogonal to common risk factors, and exploits asymmetric 
local trend and reversion patterns. Our strategies remain profitable after taking into account trading frictions and costs. Our findings suggest a high compensation for arbitrageurs to enforce the law of one price.
 

Bio:

Markus Pelger is an Assistant Professor of Management Science & Engineering at Stanford University and the Reid and Polly Anderson Faculty Fellow. His research focuses on understanding and managing financial risk. He develops mathematical financial models and statistical methods, analyzes financial data, and engineers computational techniques. His research is divided into three streams: stochastic financial modeling, high-frequency statistics, and statistical learning in high-dimensional financial data sets. His most recent work includes developing machine learning solutions to big-data problems in empirical risk management and asset pricing.
Professor Pelger is an organizer of the AI & Big Data in Finance Research Forum and of the Advanced Financial Technology Laboratories at Stanford. Professor Pelger's work has appeared in the Journal of Finance, Review of Financial Studies, Journal of Applied Probability, and Journal of Econometrics. He is an Associate Editor of Management Science and also referees for several journals in the fields of statistics, econometrics, finance, and management. He received his PhD in Economics from the University of California, Berkeley. He is a scholar of the German National Merit Foundation, and he was awarded a Fulbright Scholarship, the Institute for New Economic Thinking Prize, the Eliot J. Swan Prize, the Graduate Teaching Award at Stanford University, the Utah Winter Finance Conference Best Paper Award, the Best Paper in Asset Pricing Award at the SFS Cavalcade and the Dennis Aigner Best Paper Award of the Journal of Econometrics. He has two diplomas in Mathematics and in Economics, both with highest distinction, from the University of Bonn in Germany.

View Slides

Trading with the Momentum Transformer: An Intelligent and Interpretable Architecture

Abstract:

We introduce the Momentum Transformer, an attention-based deep learning architecture which outperforms benchmark momentum and mean-reversion trading strategies. Unlike state-of-the-art Long Short-Term Memory (LSTM) architectures, which are sequential in nature, the attention mechanism provides our architecture with a direct connection to all previous time-steps. Our architecture enables us to learn longer-term dependencies, improves performance when considering returns net of transaction costs and naturally adapts to new market regimes, such as during the SARS-CoV-2 crisis. The Momentum Transformer is inherently interpretable, providing us with greater insights into our deep learning momentum trading strategy, including how it blends different classical strategies and the past time-steps which are of the greatest significance to the model.

Bio: 

Stefan Zohren is the Deputy Director of the Oxford-Man Institute of Quantitative Finance, an Associate Professor at the Department of Engineering Science, an Associate at the Oxford Internet Institute, and a Mentor at the Creative Destruction Lab at Saïd Business School, all at the University of Oxford. He is a Fellow of the Turing Institute, the UK’s national institute for AI and data science. Stefan’s research is focused on machine learning in finance, including deep learning, reinforcement learning, network and NLP approaches, as well as early use cases of quantum computing. Outside of academia, he works as a Principal Quant at Man Group leading execution research in futures and other derivatives. Stefan is a frequent speaker on AI in finance representing the Oxford-Man Institute at academic conferences, as well as industry panels and corporate events. His work has been covered in the financial news such as Bloomberg News and Risk.

View Slides

Thematic Sentiment from Bloomberg News Topic Codes

Abstract:

In this talk, we will present a methodology to produce interpretable thematic daily aggregate sentiment from the existing overall sentiment and topic code tagging information of a given news story in Bloomberg’s Company Sentiment Data Enterprise Data Feed (EDF) product. This method avoids any supervision and ad-hoc labeling of news themes, as well as training separate thematic sentiment models to produce the result. Instead, stories are first grouped in thematic categories using a suitable unsupervised PCA / ICA (or p-ICA) model trained on the topic code co-occurrence matrix. Thereafter, thematic daily aggregate sentiment is obtained by conditioning on stories belonging to each category. By retraining the model each month (on a rolling 12-month basis), the news flow themes obtained using this method are demonstrated to be remarkably robust to changes in the topic code tagging taxonomy and fluctuations of news flow. As an application, backtesting of some thematic news sentiment equity strategies will also be discussed.

Bio:

Ivailo Dimov is a senior quant with the Quantitative Finance Research team in Bloomberg‘s CTO Office, where he provides quantitative and data science solutions to the company’s management team, as well as both internal and external clients. He has worked on both traditional derivative, risk, and alpha modeling, as well as alternative data research. At Bloomberg, he has led projects on market consensus, broker-algo selection, recommendation systems, automated news, and news topic modeling, as well as geographical alternative data. Ivailo is also an Adjunct Professor at the NYU Courant Institute of Mathematical Sciences. Prior to joining Bloomberg, Ivailo was a quant in the derivative analysis group at Goldman Sachs covering equity derivatives. He has a Ph.D. in physics from UCLA and an M.S. in mathematics in finance from NYU Courant.

View Slides