Existing graph neural networks typically rely on heuristic choices for hidden dimensions and propagation depths, which often lead to severe information loss during propagation, known as over-squashing. To address this issue, we propose Channel Capacity Constrained Estimation (C³E), a novel framework that formulates the selection of hidden dimensions and depth as a nonlinear programming problem grounded in information theory. Through modeling spectral graph neural networks as communication channels, our approach directly connects channel capacity to hidden dimensions, propagation depth, propagation mechanism, and graph structure. Extensive experiments on nine public datasets demonstrate that hidden dimensions and depths estimated by C³E can mitigate over-squashing and consistently improve representation learning. Experimental results show that over-squashing occurs due to the cumulative compression of information in representation matrices. Furthermore, our findings show that increasing hidden dimensions indeed mitigates information compression, while the role of propagation depth is more nuanced, uncovering a fundamental balance between information compression and representation complexity.

Bond markets respond differently to macroeconomic news compared to equity markets, yet most sentiment models are trained primarily on general financial or equity news data. However, bond prices often move in the opposite direction to economic optimism, making general or equity-based sentiment tools potentially misleading. We introduce BondBERT, a transformer-based language model fine-tuned on bond-specific news. BondBERT can act as the perception and reasoning component of a financial decision-support agent, providing sentiment signals that integrate with forecasting models. We propose a generalisable framework for adapting transformers to low-volatility, domain-inverse sentiment tasks by compiling and cleaning 30,000 UK bond market articles (2018-2025). BondBERT's sentiment predictions are compared against FinBERT, FinGPT, and Instruct-FinGPT using event-based correlation, up/down accuracy analyses, and LSTM forecasting across ten UK sovereign bonds. We find that BondBERT consistently produces positive correlations with bond returns, and achieves higher alignment and forecasting accuracy than the three baseline models. These results demonstrate that domain-specific sentiment adaptation better captures fixed income dynamics, bridging a gap between NLP advances and bond market analytics.

Accurate forecasting in financial markets requires integrating diverse data sources, from historical prices to macroeconomic indicators and financial news. However, existing models often fail to align these modalities effectively, limiting their practical use. In this paper, we introduce a transformer-based deep learning framework, Cross-Modal Temporal Fusion (CMTF), that fuses structured and unstructured financial data for improved market prediction. The model incorporates a tensor interpretation module for feature selection and an auto-training pipeline for efficient hyperparameter tuning. Experimental results using FTSE 100 stock data demonstrate that CMTF achieves superior performance in price direction classification compared to classical and deep learning baselines. These findings suggest that our framework is an effective and scalable solution for real-world cross-modal financial forecasting tasks.

Forecasting future stock trends remains challenging for academia and industry due to stochastic inter-stock dynamics and hierarchical intra-stock dynamics influencing stock prices.In recent years, graph neural networks have achieved remarkable performance in this problem by formulating multiple stocks as graph-structured data.However, most of these approaches rely on artificially defined factors to construct static stock graphs, which fail to capture the intrinsic interdependencies between stocks that rapidly evolve.In addition, these methods often ignore the hierarchical features of the stocks and lose distinctive information within.In this work, we propose a novel graph learning approach implemented without expert knowledge to address these issues.First, our approach automatically constructs dynamic stock graphs by entropy-driven edge generation from a signal processing perspective.Then, we further learn task-optimal dependencies between stocks via a generalized graph diffusion process on constructed stock graphs.Last, a decoupled representation learning scheme is adopted to capture distinctive hierarchical intra-stock features.Experimental results demonstrate substantial improvements over state-of-the-art baselines on real-world datasets.Moreover, the ablation study and sensitivity study further illustrate the effectiveness of the proposed method in modeling the time-evolving inter-stock and intra-stock dynamics.

Stock trend classification remains a fundamental yet challenging task, owing to the intricate time-evolving dynamics between and within stocks. To tackle these two challenges, we propose a graph-based representation learning approach aimed at predicting the future movements of multiple stocks. Initially, we model the complex time-varying relationships between stocks by generating dynamic multi-relational stock graphs. This is achieved through a novel edge generation algorithm that leverages information entropy and signal energy to quantify the intensity and directionality of inter-stock relations on each trading day. Then, we further refine these initial graphs through a stochastic multi-relational diffusion process, adaptively learning task-optimal edges. Subsequently, we implement a decoupled representation learning scheme with parallel retention to obtain the final graph representation. This strategy better captures the unique temporal features within individual stocks while also capturing the overall structure of the stock graph. Comprehensive experiments conducted on real-world datasets from two US markets (NASDAQ and NYSE) and one Chinese market (Shanghai Stock Exchange: SSE) validate the effectiveness of our method. Our approach1consistently outperforms state-of-the-art baselines in forecasting next trading day stock trends across three test periods spanning seven years.

This research explores the utility of ontological representations using object-oriented (OO) design principles, such that characteristics of the problem domain are directly mapped onto the representation of individuals. A comparison against more traditional representations is performed in two problem domains of differing complexity: (i) Tangram, a simple geometric puzzle; and (ii) EvoRecSys, an evolutionary recommender system for health and well-being advice. We show that OO representations aid research and development as naturally decoupled components can be more easily modified and extended, which can in turn lead to the discovery of better solutions.

The majority of trading in financial markets is executed through a limit order book (LOB). The LOB is an event-based continuously-updating system that records contemporaneous demand (`bids' to buy) and supply (`asks' to sell) for a financial asset. Following recent successes in the literature that combine stochastic point processes with neural networks to model event stream patterns, we propose a novel state-dependent parallel neural Hawkes process to predict LOB events and simulate realistic LOB data. The model is characterized by: (1) separate intensity rate modelling for each event type through a parallel structure of continuous time LSTM units; and (2) an event-state interaction mechanism that improves prediction accuracy and enables efficient sampling of the event-state stream. We first demonstrate the superiority of the proposed model over traditional stochastic or deep learning models for predicting event type and time of a real world LOB dataset. Using stochastic point sampling from a well trained model, we then develop a realistic deep learning-based LOB simulator that exhibits multiple stylized facts found in real LOB data.

Abstract In recent years, recommender systems have been employed in domains like e-commerce, tourism, and multimedia streaming, where personalising users’ experience based on their interactions is a fundamental aspect to consider. Recent recommender system developments have also focused on well-being, yet existing solutions have been entirely designed considering one single well-being aspect in isolation, such as a healthy diet or an active lifestyle. This research introduces EvoRecSys, a novel recommendation framework that proposes evolutionary algorithms as the main recommendation engine, thereby modelling the problem of generating personalised well-being recommendations as a multi-objective optimisation problem. EvoRecSys captures the interrelation between multiple aspects of well-being by constructing configurable recommendations in the form of bundled items with dynamic properties. The preferences and a predefined well-being goal by the user are jointly considered. By instantiating the framework into an implemented model, we illustrate the use of a genetic algorithm as the recommendation engine. Finally, this implementation has been deployed as a Web application in order to conduct a users’ study.

In an order-driven financial market, the price of a financial asset is discovered through the interaction of orders - requests to buy or sell at a particular price - that are posted to the public limit order book (LOB). Therefore, LOB data is extremely valuable for modelling market dynamics. However, LOB data is not freely accessible, which poses a challenge to market participants and researchers wishing to exploit this information. Fortunately, trades and quotes (TAQ) data - orders arriving at the top of the LOB, and trades executing in the market - are more readily available. In this paper, we present the LOB recreation model, a first attempt from a deep learning perspective to recreate the top five price levels of the LOB for small-tick stocks using only TAQ data. Volumes of orders sitting deep in the LOB are predicted by combining outputs from: (1) a history compiler that uses a Gated Recurrent Unit (GRU) module to selectively compile prediction relevant quote history; (2) a market events simulator, which uses an Ordinary Differential Equation Recurrent Neural Network (ODE-RNN) to simulate the accumulation of net order arrivals; and (3) a weighting scheme to adaptively combine the predictions generated by (1) and (2). By the paradigm of transfer learning, the core encoder trained on one stock can be fine-tuned to enable application to other financial assets of the same class with much lower demand on additional data. Comprehensive experiments conducted on two real world intraday LOB datasets demonstrate that the proposed model can efficiently recreate the LOB with high accuracy using only TAQ data as input.

We propose substitution of the fittest (SF), a novel technique designed to counteract the problem of disengagement in two-population competitive coevolutionary genetic algorithms. The approach presented is domain-independent and requires no calibration. In a minimal domain, we perform a controlled evaluation of the ability to maintain engagement and the capacity to discover optimal solutions. Results demonstrate that the solution discovery performance of SF is comparable with other techniques in the literature, while SF also offers benefits including a greater ability to maintain engagement and a much simpler mechanism.

We explore the competitive effects of reaction time of automated trading strategies in simulated financial markets containing a single exchange with public limit order book and continuous double auction matching. A large body of research conducted over several decades has been devoted to trading agent design and simulation, but the majority of this work focuses on pricing strategy and does not consider the time taken for these strategies to compute. In real-world financial markets, speed is known to heavily influence the design of automated trading algorithms, with the generally accepted wisdom that faster is better. Here, we introduce increasingly realistic models of trading speed and profile the computation times of a suite of eminent trading algorithms from the literature. Results demonstrate that: (a) trading performance is impacted by speed, but faster is not always better; (b) the Adaptive-Aggressive (AA) algorithm, until recently considered the most dominant trading strategy in the literature, is outperformed by the simplistic Shaver (SHVR) strategy - shave one tick off the current best bid or ask - when relative computation times are accurately simulated.

Global positioning system (GPS) data generated from taxi trips is a valuable source of information that offers an insight into travel behaviours of urban populations with high spatio-temporal resolution. However, in its raw form, GPS taxi data does not offer information on the purpose (or intended activity) of travel. In this context, to enhance the utility of taxi GPS data sets, we propose a two-layer framework to identify the related activities of each taxi trip automatically and estimate the return trips and successive activities after the trip, by using geographic point-of-interest (POI) data and a combination of spatio-temporal clustering, Bayesian inference and Monte Carlo simulation. Two million taxi trips in New York, the United States of America, and ten million taxi trips in Shenzhen, China, are used as inputs for the two-layer framework. To validate each layer of the framework, we collect 6,003 trip diaries in New York and 712 questionnaire surveys in Shenzhen. The results show that the first layer of the framework performs better than comparable methods published in the literature, while the second layer has high accuracy when inferring return trips.

We consider the issue of securing dark pools/markets in thefinancial services sector. Currently, these markets are executed via trusted third parties, which opens potential for market operators to conduct fraud. We present a potential solution to this problem by using Multi-Party Computation (MPC) to enable a trusted third party to be emulated in software for three popular market mechanisms: continuous double auction (CDA), periodic auction (PA), and scheduled volume matching (SVM). Our experiments show that whilst the predominate market clearing mechanism of CDA is not currently viable when executed using MPC, SVM (a popular market mechanism for dark pools) is viable. We present experimental validation of this conclusion by presenting the expected throughputs for such markets in two popular MPC paradigms: the two party dishonest majority setting, and the honest majority three party setting.

In September 2012, Amazon, the leading Infrastructure as a Service (IaaS) provider, launched a secondary marketplace venue for users to buy and sell cloud resources between themselves—the Amazon EC2 Reserved Instance Marketplace (ARIM). ARIM is designed to encourage users to purchase more long-term reserved instances, thus generating more stable demand for the provider and additional revenue through commission on sales. In this paper, we model ARIM using a multi-agent simulation model populated with zero-intelligence plus (ZIP) financial trading agents. We demonstrate that ARIM offers a new opportunity for market makers (MMs) to profit from buying and selling resources, but suggest that this opportunity may be fleeting. We also demonstrate that altering the market mechanism from a retail market (where only sellers post offers; similar to ARIM) to a continuous double auction (where both buyers and sellers post offers) can result in higher sale prices and therefore higher commissions. Since IaaS is a multi-billion dollar industry and currently the fastest growing segment of the cloud computing market, we therefore suggest that Amazon may profit from altering the mechanism of ARIM to enable buyers to post bids.

A novel approach for the autonomous virulence adaptation (AVA) of competing populations in a coevolutionary optimization framework is presented. Previous work has demonstrated that setting an appropriate virulence,$v$, of populations accelerates coevolutionary optimization by avoiding detrimental periods of disengagement. However, since the likelihood of disengagement varies both between systems and over time, choosing the ideal value of$v$is problematic. The AVA technique presented here uses a machine learning approach to continuously tune$v$as system engagement varies. In a simple, abstract domain, AVA is shown to successfully adapt to the most productive values of$v$. Further experiments, in more complex domains of sorting networks and maze navigation, demonstrate AVA's efficiency over reduced virulence and the layered Pareto coevolutionary archive.

While standard evolutionary algorithms employ a static, absolute fitness metric, coevolutionary algorithms assess individuals by their performance relative to populations of opponents that are themselves evolving. Although this arrangement offers the possibility of avoiding long-standing difficulties such as premature convergence, it suffers from its own unique problems, cycling, over-focusing and disengagement. Here, we introduce a novel technique for dealing with the third and least explored of these problems. Inspired by studies of natural host-parasite systems, we show that disengagement can be avoided by selecting for individuals that exhibit reduced levels of "virulence", rather than maximum ability to defeat coevolutionary adversaries. Experiments in both simple and complex domains are used to explain how this counterintuitive approach may be used to improve the success of coevolutionary algorithms.

Inspired by the virulence of natural parasites, a novel approach is developed to tackle disengagement, a detrimental phenomenon coevolutionary systems sometimes experience. After demonstrating beneficial results in a simple model, minimum comparison sorting networks are coevolved, with results suggesting that moderating parasite virulence can help in practical problem domains.