Earnings Prediction Using Recurrent Neural Networks ArXiv ID: 2311.10756 “View on arXiv” Authors: Unknown Abstract Firm disclosures about future prospects are crucial for corporate valuation and compliance with global regulations, such as the EU’s MAR and the US’s SEC Rule 10b-5 and RegFD. To comply with disclosure obligations, issuers must identify nonpublic information with potential material impact on security prices as only new, relevant and unexpected information materially affects prices in efficient markets. Financial analysts, assumed to represent public knowledge on firms’ earnings prospects, face limitations in offering comprehensive coverage and unbiased estimates. This study develops a neural network to forecast future firm earnings, using four decades of financial data, addressing analysts’ coverage gaps and potentially revealing hidden insights. The model avoids selectivity and survivorship biases as it allows for missing data. Furthermore, the model is able to produce both fiscal-year-end and quarterly earnings predictions. Its performance surpasses benchmark models from the academic literature by a wide margin and outperforms analysts’ forecasts for fiscal-year-end earnings predictions. ...

Enhancing Actuarial Non-Life Pricing Models via Transformers ArXiv ID: 2311.07597 “View on arXiv” Authors: Unknown Abstract Currently, there is a lot of research in the field of neural networks for non-life insurance pricing. The usual goal is to improve the predictive power via neural networks while building upon the generalized linear model, which is the current industry standard. Our paper contributes to this current journey via novel methods to enhance actuarial non-life models with transformer models for tabular data. We build here upon the foundation laid out by the combined actuarial neural network as well as the localGLMnet and enhance those models via the feature tokenizer transformer. The manuscript demonstrates the performance of the proposed methods on a real-world claim frequency dataset and compares them with several benchmark models such as generalized linear models, feed-forward neural networks, combined actuarial neural networks, LocalGLMnet, and pure feature tokenizer transformer. The paper shows that the new methods can achieve better results than the benchmark models while preserving certain generalized linear model advantages. The paper also discusses the practical implications and challenges of applying transformer models in actuarial settings. ...

Multi-Label Topic Model for Financial Textual Data ArXiv ID: 2311.07598 “View on arXiv” Authors: Unknown Abstract This paper presents a multi-label topic model for financial texts like ad-hoc announcements, 8-K filings, finance related news or annual reports. I train the model on a new financial multi-label database consisting of 3,044 German ad-hoc announcements that are labeled manually using 20 predefined, economically motivated topics. The best model achieves a macro F1 score of more than 85%. Translating the data results in an English version of the model with similar performance. As application of the model, I investigate differences in stock market reactions across topics. I find evidence for strong positive or negative market reactions for some topics, like announcements of new Large Scale Projects or Bankruptcy Filings, while I do not observe significant price effects for some other topics. Furthermore, in contrast to previous studies, the multi-label structure of the model allows to analyze the effects of co-occurring topics on stock market reactions. For many cases, the reaction to a specific topic depends heavily on the co-occurrence with other topics. For example, if allocated capital from a Seasoned Equity Offering (SEO) is used for restructuring a company in the course of a Bankruptcy Proceeding, the market reacts positively on average. However, if that capital is used for covering unexpected, additional costs from the development of new drugs, the SEO implies negative reactions on average. ...

Advancing Algorithmic Trading: A Multi-Technique Enhancement of Deep Q-Network Models ArXiv ID: 2311.05743 “View on arXiv” Authors: Unknown Abstract This study enhances a Deep Q-Network (DQN) trading model by incorporating advanced techniques like Prioritized Experience Replay, Regularized Q-Learning, Noisy Networks, Dueling, and Double DQN. Extensive tests on assets like BTC/USD and AAPL demonstrate superior performance compared to the original model, with marked increases in returns and Sharpe Ratio, indicating improved risk-adjusted rewards. Notably, convolutional neural network (CNN) architectures, both 1D and 2D, significantly boost returns, suggesting their effectiveness in market trend analysis. Across instruments, these enhancements have yielded stable and high gains, eclipsing the baseline and highlighting the potential of CNNs in trading systems. The study suggests that applying sophisticated deep learning within reinforcement learning can greatly enhance automated trading, urging further exploration into advanced methods for broader financial applicability. The findings advocate for the continued evolution of AI in finance. ...

Optimal dividend strategies for a catastrophe insurer ArXiv ID: 2311.05781 “View on arXiv” Authors: Unknown Abstract In this paper we study the problem of optimally paying out dividends from an insurance portfolio, when the criterion is to maximize the expected discounted dividends over the lifetime of the company and the portfolio contains claims due to natural catastrophes, modelled by a shot-noise Cox claim number process. The optimal value function of the resulting two-dimensional stochastic control problem is shown to be the smallest viscosity supersolution of a corresponding Hamilton-Jacobi-Bellman equation, and we prove that it can be uniformly approximated through a discretization of the space of the free surplus of the portfolio and the current claim intensity level. We implement the resulting numerical scheme to identify optimal dividend strategies for such a natural catastrophe insurer, and it is shown that the nature of the barrier and band strategies known from the classical models with constant Poisson claim intensity carry over in a certain way to this more general situation, leading to action and non-action regions for the dividend payments as a function of the current surplus and intensity level. We also discuss some interpretations in terms of upward potential for shareholders when including a catastrophe sector in the portfolio. ...

Causal Inference on Investment Constraints and Non-stationarity in Dynamic Portfolio Optimization through Reinforcement Learning ArXiv ID: 2311.04946 “View on arXiv” Authors: Unknown Abstract In this study, we have developed a dynamic asset allocation investment strategy using reinforcement learning techniques. To begin with, we have addressed the crucial issue of incorporating non-stationarity of financial time series data into reinforcement learning algorithms, which is a significant implementation in the application of reinforcement learning in investment strategies. Our findings highlight the significance of introducing certain variables such as regime change in the environment setting to enhance the prediction accuracy. Furthermore, the application of reinforcement learning in investment strategies provides a remarkable advantage of setting the optimization problem flexibly. This enables the integration of practical constraints faced by investors into the algorithm, resulting in efficient optimization. Our study has categorized the investment strategy formulation conditions into three main categories, including performance measurement indicators, portfolio management rules, and other constraints. We have evaluated the impact of incorporating these conditions into the environment and rewards in a reinforcement learning framework and examined how they influence investment behavior. ...

Explainable artificial intelligence model for identifying Market Value in Professional Soccer Players ArXiv ID: 2311.04599 “View on arXiv” Authors: Unknown Abstract This study introduces an advanced machine learning method for predicting soccer players’ market values, combining ensemble models and the Shapley Additive Explanations (SHAP) for interpretability. Utilizing data from about 12,000 players from Sofifa, the Boruta algorithm streamlined feature selection. The Gradient Boosting Decision Tree (GBDT) model excelled in predictive accuracy, with an R-squared of 0.901 and a Root Mean Squared Error (RMSE) of 3,221,632.175. Player attributes in skills, fitness, and cognitive areas significantly influenced market value. These insights aid sports industry stakeholders in player valuation. However, the study has limitations, like underestimating superstar players’ values and needing larger datasets. Future research directions include enhancing the model’s applicability and exploring value prediction in various contexts. ...

Forecasting Volatility with Machine Learning and Rough Volatility: Example from the Crypto-Winter ArXiv ID: 2311.04727 “View on arXiv” Authors: Unknown Abstract We extend the application and test the performance of a recently introduced volatility prediction framework encompassing LSTM and rough volatility. Our asset class of interest is cryptocurrencies, at the beginning of the “crypto-winter” in 2022. We first show that to forecast volatility, a universal LSTM approach trained on a pool of assets outperforms traditional models. We then consider a parsimonious parametric model based on rough volatility and Zumbach effect. We obtain similar prediction performances with only five parameters whose values are non-asset-dependent. Our findings provide further evidence on the universality of the mechanisms underlying the volatility formation process. ...

Portfolio Construction using Black-Litterman Model and Factors ArXiv ID: 2311.04475 “View on arXiv” Authors: Unknown Abstract This paper presents a portfolio construction process, including mainly two parts, Factors Selection and Weight Allocations. For the factors selection part, We have chosen 20 factors by considering three aspects, the global market, different assets class, and stock idiosyncratic characteristics. Each factor is proxied by a corresponding ETF. Then, we would apply several weight allocation methods to those factors, including two fixed weight allocation methods, three optimisation methods, and a Black-Litterman model. In addition, we would also fit a Deep Learning model for generating views periodically and incorporating views with the prior to achieve dynamically updated weights by using the Black-Litterman model. In the end, the robustness checking shows how weights change with respect to time evolving and variance increasing. Results using shrinkage variance are provided to alleviate the impacts of representativeness of historical data, but there sadly has little impact. Overall, the model by using the Deep Learning plus Black-Litterman model results outperform the portfolio by other weight allocation schemes, even though further improvement and robustness checking should be performed. ...

On an Optimal Stopping Problem with a Discontinuous Reward ArXiv ID: 2311.03538 “View on arXiv” Authors: Unknown Abstract We study an optimal stopping problem with an unbounded, time-dependent and discontinuous reward function. This problem is motivated by the pricing of a variable annuity contract with guaranteed minimum maturity benefit, under the assumption that the policyholder’s surrender behaviour maximizes the risk-neutral value of the contract. We consider a general fee and surrender charge function, and give a condition under which optimal stopping always occurs at maturity. Using an alternative representation for the value function of the optimization problem, we study its analytical properties and the resulting surrender (or exercise) region. In particular, we show that the non-emptiness and the shape of the surrender region are fully characterized by the fee and the surrender charge functions, which provides a powerful tool to understand their interrelation and how it affects early surrenders and the optimal surrender boundary. Under certain conditions on these two functions, we develop three representations for the value function; two are analogous to their American option counterpart, and one is new to the actuarial and American option pricing literature. ...