Stephan Robert-Nicoud

Denis Levchenko, Efstratios Rappos, Shabnam Ataee et al.

Neural architecture search (NAS) emerged as a way to automatically optimize neural networks for a specific task and dataset. Despite an abundance of research on NAS for images and natural language applications, similar studies for time series data are lacking. Among NAS search spaces, chain-structured are the simplest and most applicable to small datasets like time series. We compare three popular NAS strategies on chain-structured search spaces: Bayesian optimization (specifically Tree-structured Parzen Estimator), the hyperband method, and reinforcement learning in the context of financial time series forecasting. These strategies were employed to optimize simple well-understood neural architectures like the MLP, 1D CNN, and RNN, with more complex temporal fusion transformers (TFT) and their own optimizers included for comparison. We find Bayesian optimization and the hyperband method performing best among the strategies, and RNN and 1D CNN best among the architectures, but all methods were very close to each other with a high variance due to the difficulty of working with financial datasets. We discuss our approach to overcome the variance and provide implementation recommendations for future users and researchers.

Félicien Hêche, Biagio Nigro, Oussama Barakat et al.

Financial markets have experienced significant instabilities in recent years, creating unique challenges for trading and increasing interest in risk-averse strategies. Distributional Reinforcement Learning (RL) algorithms, which model the full distribution of returns rather than just expected values, offer a promising approach to managing market uncertainty. This paper investigates this potential by studying the effectiveness of three distributional RL algorithms for natural gas futures trading and exploring their capacity to develop risk-averse policies. Specifically, we analyze the performance and behavior of Categorical Deep Q-Network (C51), Quantile Regression Deep Q-Network (QR-DQN), and Implicit Quantile Network (IQN). To the best of our knowledge, these algorithms have never been applied in a trading context. These policies are compared against five Machine Learning (ML) baselines, using a detailed dataset provided by Predictive Layer SA, a company supplying ML-based strategies for energy trading. The main contributions of this study are as follows. (1) We demonstrate that distributional RL algorithms significantly outperform classical RL methods, with C51 achieving performance improvement of more than 32\%. (2) We show that training C51 and IQN to maximize CVaR produces risk-sensitive policies with adjustable risk aversion. Specifically, our ablation studies reveal that lower CVaR confidence levels increase risk aversion, while higher levels decrease it, offering flexible risk management options. In contrast, QR-DQN shows less predictable behavior. These findings emphasize the potential of distributional RL for developing adaptable, risk-averse trading strategies in volatile markets.

Regis Houssou, Mihai-Cezar Augustin, Efstratios Rappos et al.

This paper proposes a new method to generate synthetic data sets based on copula models. Our goal is to produce surrogate data resembling real data in terms of marginal and joint distributions. We present a complete and reliable algorithm for generating a synthetic data set comprising numeric or categorical variables. Applying our methodology to two datasets shows better performance compared to other methods such as SMOTE and autoencoders.

Mihai-Cezar Augustin, Vivien Bonvin, Regis Houssou et al.

In classification problems, supervised machine-learning methods outperform traditional algorithms, thanks to the ability of neural networks to learn complex patterns. However, in two-class classification tasks like anomaly or fraud detection, unsupervised methods could do even better, because their prediction is not limited to previously learned types of anomalies. An intuitive approach of anomaly detection can be based on the distances from the centers of mass of the two respective classes. Autoencoders, although trained without supervision, can also detect anomalies: considering the center of mass of the normal points, reconstructions have now radii, with largest radii most likely indicating anomalous points. Of course, radii-based classification were already possible without interposing an autoencoder. In any space, radial classification can be operated, to some extent. In order to outperform it, we proceed to radial deformations of data (i.e. centric compression or expansions of axes) and autoencoder training. Any autoencoder that makes use of a data center is here baptized a centric autoencoder (cAE). A special type is the cAE trained with a uniformly compressed dataset, named the centripetal autoencoder (cpAE). The new concept is studied here in relation with a schematic artificial dataset, and the derived methods show consistent score improvements. But tested on real banking data, our radial deformation supervised algorithms alone still perform better that cAEs, as expected from most supervised methods; nonetheless, in hybrid approaches, cAEs can be combined with a radial deformation of space, improving its classification score. We expect that centric autoencoders will become irreplaceable objects in anomaly live detection based on geometry, thanks to their ability to stem naturally on geometrical algorithms and to their native capability of detecting unknown anomaly types.

Giulia Moschini, Régis Houssou, Jérôme Bovay et al.

This paper addresses the problem of unsupervised approach of credit card fraud detection in unbalanced dataset using the ARIMA model. The ARIMA model is fitted on the regular spending behaviour of the customer and is used to detect fraud if some deviations or discrepancies appear. Our model is applied to credit card datasets and is compared to 4 anomaly detection approaches such as K-Means, Box-Plot, Local Outlier Factor and Isolation Forest. The results show that the ARIMA model presents a better detecting power than the benchmark models.