Deep Learning modeling of Limit Order Book: a comparative perspective

TL;DR

This paper compares various deep learning models for high-frequency trading based on Limit Order Book data, analyzing their performance and underlying dynamics to understand which models best capture market behavior.

Contribution

It provides a comprehensive comparison of state-of-the-art deep learning models on the same trading dataset, revealing insights into the intrinsic dimensions of Limit Order Book dynamics.

Findings

MLP performs comparably or better than CNN-LSTM models.

Dynamic spatial and temporal dimensions approximate LOB dynamics.

Intrinsic dimensions of LOB are not fully captured by current models.

Abstract

The present work addresses theoretical and practical questions in the domain of Deep Learning for High Frequency Trading. State-of-the-art models such as Random models, Logistic Regressions, LSTMs, LSTMs equipped with an Attention mask, CNN-LSTMs and MLPs are reviewed and compared on the same tasks, feature space and dataset, and then clustered according to pairwise similarity and performance metrics. The underlying dimensions of the modeling techniques are hence investigated to understand whether these are intrinsic to the Limit Order Book's dynamics. We observe that the Multilayer Perceptron performs comparably to or better than state-of-the-art CNN-LSTM architectures indicating that dynamic spatial and temporal dimensions are a good approximation of the LOB's dynamics, but not necessarily the true underlying dimensions.