Large (and Deep) Factor Models ArXiv ID: 2402.06635 “View on arXiv” Authors: Unknown Abstract We open up the black box behind Deep Learning for portfolio optimization and prove that a sufficiently wide and arbitrarily deep neural network (DNN) trained to maximize the Sharpe ratio of the Stochastic Discount Factor (SDF) is equivalent to a large factor model (LFM): A linear factor pricing model that uses many non-linear characteristics. The nature of these characteristics depends on the architecture of the DNN in an explicit, tractable fashion. This makes it possible to derive end-to-end trained DNN-based SDFs in closed form for the first time. We evaluate LFMs empirically and show how various architectural choices impact SDF performance. We document the virtue of depth complexity: With enough data, the out-of-sample performance of DNN-SDF is increasing in the NN depth, saturating at huge depths of around 100 hidden layers. ...

BioFinBERT: Finetuning Large Language Models (LLMs) to Analyze Sentiment of Press Releases and Financial Text Around Inflection Points of Biotech Stocks ArXiv ID: 2401.11011 “View on arXiv” Authors: Unknown Abstract Large language models (LLMs) are deep learning algorithms being used to perform natural language processing tasks in various fields, from social sciences to finance and biomedical sciences. Developing and training a new LLM can be very computationally expensive, so it is becoming a common practice to take existing LLMs and finetune them with carefully curated datasets for desired applications in different fields. Here, we present BioFinBERT, a finetuned LLM to perform financial sentiment analysis of public text associated with stocks of companies in the biotechnology sector. The stocks of biotech companies developing highly innovative and risky therapeutic drugs tend to respond very positively or negatively upon a successful or failed clinical readout or regulatory approval of their drug, respectively. These clinical or regulatory results are disclosed by the biotech companies via press releases, which are followed by a significant stock response in many cases. In our attempt to design a LLM capable of analyzing the sentiment of these press releases,we first finetuned BioBERT, a biomedical language representation model designed for biomedical text mining, using financial textual databases. Our finetuned model, termed BioFinBERT, was then used to perform financial sentiment analysis of various biotech-related press releases and financial text around inflection points that significantly affected the price of biotech stocks. ...

MDGNN: Multi-Relational Dynamic Graph Neural Network for Comprehensive and Dynamic Stock Investment Prediction ArXiv ID: 2402.06633 “View on arXiv” Authors: Unknown Abstract The stock market is a crucial component of the financial system, but predicting the movement of stock prices is challenging due to the dynamic and intricate relations arising from various aspects such as economic indicators, financial reports, global news, and investor sentiment. Traditional sequential methods and graph-based models have been applied in stock movement prediction, but they have limitations in capturing the multifaceted and temporal influences in stock price movements. To address these challenges, the Multi-relational Dynamic Graph Neural Network (MDGNN) framework is proposed, which utilizes a discrete dynamic graph to comprehensively capture multifaceted relations among stocks and their evolution over time. The representation generated from the graph offers a complete perspective on the interrelationships among stocks and associated entities. Additionally, the power of the Transformer structure is leveraged to encode the temporal evolution of multiplex relations, providing a dynamic and effective approach to predicting stock investment. Further, our proposed MDGNN framework achieves the best performance in public datasets compared with state-of-the-art (SOTA) stock investment methods. ...

Stylized Facts and Market Microstructure: An In-Depth Exploration of German Bond Futures Market ArXiv ID: 2401.10722 “View on arXiv” Authors: Unknown Abstract This paper presents an in-depth analysis of stylized facts in the context of futures on German bonds. The study examines four futures contracts on German bonds: Schatz, Bobl, Bund and Buxl, using tick-by-tick limit order book datasets. It uncovers a range of stylized facts and empirical observations, including the distribution of order sizes, patterns of order flow, and inter-arrival times of orders. The findings reveal both commonalities and unique characteristics across the different futures, thereby enriching our understanding of these markets. Furthermore, the paper introduces insightful realism metrics that can be used to benchmark market simulators. The study contributes to the literature on financial stylized facts by extending empirical observations to this class of assets, which has been relatively underexplored in existing research. This work provides valuable guidance for the development of more accurate and realistic market simulators. ...

Consistent asset modelling with random coefficients and switches between regimes ArXiv ID: 2401.09955 “View on arXiv” Authors: Unknown Abstract We explore a stochastic model that enables capturing external influences in two specific ways. The model allows for the expression of uncertainty in the parametrisation of the stochastic dynamics and incorporates patterns to account for different behaviours across various times or regimes. To establish our framework, we initially construct a model with random parameters, where the switching between regimes can be dictated either by random variables or deterministically. Such a model is highly interpretable. We further ensure mathematical consistency by demonstrating that the framework can be elegantly expressed through local volatility models taking the form of standard jump diffusions. Additionally, we consider a Markov-modulated approach for the switching between regimes characterised by random parameters. For all considered models, we derive characteristic functions, providing a versatile tool with wide-ranging applications. In a numerical experiment, we apply the framework to the financial problem of option pricing. The impact of parameter uncertainty is analysed in a two-regime model, where the asset process switches between periods of high and low volatility imbued with high and low uncertainty, respectively. ...

Cross-Domain Behavioral Credit Modeling: transferability from private to central data ArXiv ID: 2401.09778 “View on arXiv” Authors: Unknown Abstract This paper introduces a credit risk rating model for credit risk assessment in quantitative finance, aiming to categorize borrowers based on their behavioral data. The model is trained on data from Experian, a widely recognized credit bureau, to effectively identify instances of loan defaults among bank customers. Employing state-of-the-art statistical and machine learning techniques ensures the model’s predictive accuracy. Furthermore, we assess the model’s transferability by testing it on behavioral data from the Bank of Italy, demonstrating its potential applicability across diverse datasets during prediction. This study highlights the benefits of incorporating external behavioral data to improve credit risk assessment in financial institutions. ...

Deep Generative Modeling for Financial Time Series with Application in VaR: A Comparative Review ArXiv ID: 2401.10370 “View on arXiv” Authors: Unknown Abstract In the financial services industry, forecasting the risk factor distribution conditional on the history and the current market environment is the key to market risk modeling in general and value at risk (VaR) model in particular. As one of the most widely adopted VaR models in commercial banks, Historical simulation (HS) uses the empirical distribution of daily returns in a historical window as the forecast distribution of risk factor returns in the next day. The objectives for financial time series generation are to generate synthetic data paths with good variety, and similar distribution and dynamics to the original historical data. In this paper, we apply multiple existing deep generative methods (e.g., CGAN, CWGAN, Diffusion, and Signature WGAN) for conditional time series generation, and propose and test two new methods for conditional multi-step time series generation, namely Encoder-Decoder CGAN and Conditional TimeVAE. Furthermore, we introduce a comprehensive framework with a set of KPIs to measure the quality of the generated time series for financial modeling. The KPIs cover distribution distance, autocorrelation and backtesting. All models (HS, parametric and neural networks) are tested on both historical USD yield curve data and additional data simulated from GARCH and CIR processes. The study shows that top performing models are HS, GARCH and CWGAN models. Future research directions in this area are also discussed. ...

A closer look at the chemical potential of an ideal agent system ArXiv ID: 2401.09233 “View on arXiv” Authors: Unknown Abstract Models for spin systems known from statistical physics are used in econometrics in the form of agent-based models. Econophysics research in econometrics is increasingly developing general market models that describe exchange phenomena and use the chemical potential $μ$ known from physics in the context of particle number changes. In statistical physics, equations of state are known for the chemical potential, which take into account the respective model framework and the corresponding state variables. A simple transfer of these equations of state to problems in econophysics appears difficult. To the best of our knowledge, the equation of state for the chemical potential is currently missing even for the simplest conceivable model of an ideal agent system. In this paper, this research gap is closed and the equation of state for the chemical potential is derived from the econophysical model assumptions of the ideal agent system. An interpretation of the equation of state leads to fundamental relationships that could also have been guessed, but are shown here by the theory. ...

Neural Hawkes: Non-Parametric Estimation in High Dimension and Causality Analysis in Cryptocurrency Markets ArXiv ID: 2401.09361 “View on arXiv” Authors: Unknown Abstract We propose a novel approach to marked Hawkes kernel inference which we name the moment-based neural Hawkes estimation method. Hawkes processes are fully characterized by their first and second order statistics through a Fredholm integral equation of the second kind. Using recent advances in solving partial differential equations with physics-informed neural networks, we provide a numerical procedure to solve this integral equation in high dimension. Together with an adapted training pipeline, we give a generic set of hyperparameters that produces robust results across a wide range of kernel shapes. We conduct an extensive numerical validation on simulated data. We finally propose two applications of the method to the analysis of the microstructure of cryptocurrency markets. In a first application we extract the influence of volume on the arrival rate of BTC-USD trades and in a second application we analyze the causality relationships and their directions amongst a universe of 15 cryptocurrency pairs in a centralized exchange. ...

A Two-Step Longstaff Schwartz Monte Carlo Approach to Game Option Pricing ArXiv ID: 2401.08093 “View on arXiv” Authors: Unknown Abstract We proposed a two-step Longstaff Schwartz Monte Carlo (LSMC) method with two regression models fitted at each time step to price game options. Although the original LSMC can be used to price game options with an enlarged range of path in regression and a modified cashflow updating rule, we identified a drawback of such approach, which motivated us to propose our approach. We implemented numerical examples with benchmarks using binomial tree and numerical PDE, and it showed that our method produces more reliable results comparing to the original LSMC. ...