Neural Stochastic Agent-Based Limit Order Book Simulation: A Hybrid Methodology

Realistic limit order book (LOB) simulations are essential in understanding market dynamics. Mainstream simulation models include agent-based models (ABMs) and stochastic models (SMs). However, ABMs tend not to be grounded on real historical data, while SMs tend not to enable dynamic LOB interaction. Here, we propose a hybrid LOB simulation paradigm characterised by: (1) representing the aggregation of market events' logic by a neural stochastic background trader (BT) that is pre-trained on historical LOB data through a neural point process model; and (2) embedding the BT into an ABM to enable responsive interaction. Empirical results demonstrate that system behaviours exhibit multiple stylised facts, and the results of interaction between the BT and various trading strategies are in accordance with observations of real markets. CCS CONCEPTS • Computing methodologies → Agent / discrete models; Neural networks; • Applied computing → Economics.