Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Predicting Price Movements in High-Frequency Financial Data with Spiking Neural Networks

Brian Ezinwoke, Oliver Rhodes

TL;DR

The paper tackles price-spike forecasting in high-frequency financial data where traditional models struggle with millisecond-scale events. It deploys three Spiking Neural Network architectures trained with Spike-Timing-Dependent Plasticity, guided by Bayesian optimization using a novel Penalised Spike Accuracy objective. PSA-optimised unsupervised models, especially the extended inhibitory architecture, deliver superior predictive and trading performance compared with a supervised SNN baseline. The study provides a reproducible end-to-end framework and demonstrates the viability of STDP-trained SNNs for HFT spike forecasting and momentum trading, with implications for neuromorphic deployment.

Abstract

Modern high-frequency trading (HFT) environments are characterized by sudden price spikes that present both risk and opportunity, but conventional financial models often fail to capture the required fine temporal structure. Spiking Neural Networks (SNNs) offer a biologically inspired framework well-suited to these challenges due to their natural ability to process discrete events and preserve millisecond-scale timing. This work investigates the application of SNNs to high-frequency price-spike forecasting, enhancing performance via robust hyperparameter tuning with Bayesian Optimization (BO). This work converts high-frequency stock data into spike trains and evaluates three architectures: an established unsupervised STDP-trained SNN, a novel SNN with explicit inhibitory competition, and a supervised backpropagation network. BO was driven by a novel objective, Penalized Spike Accuracy (PSA), designed to ensure a network's predicted price spike rate aligns with the empirical rate of price events. Simulated trading demonstrated that models optimized with PSA consistently outperformed their Spike Accuracy (SA)-tuned counterparts and baselines. Specifically, the extended SNN model with PSA achieved the highest cumulative return (76.8%) in simple backtesting, significantly surpassing the supervised alternative (42.54% return). These results validate the potential of spiking networks, when robustly tuned with task-specific objectives, for effective price spike forecasting in HFT.

Predicting Price Movements in High-Frequency Financial Data with Spiking Neural Networks

TL;DR

The paper tackles price-spike forecasting in high-frequency financial data where traditional models struggle with millisecond-scale events. It deploys three Spiking Neural Network architectures trained with Spike-Timing-Dependent Plasticity, guided by Bayesian optimization using a novel Penalised Spike Accuracy objective. PSA-optimised unsupervised models, especially the extended inhibitory architecture, deliver superior predictive and trading performance compared with a supervised SNN baseline. The study provides a reproducible end-to-end framework and demonstrates the viability of STDP-trained SNNs for HFT spike forecasting and momentum trading, with implications for neuromorphic deployment.

Abstract

Modern high-frequency trading (HFT) environments are characterized by sudden price spikes that present both risk and opportunity, but conventional financial models often fail to capture the required fine temporal structure. Spiking Neural Networks (SNNs) offer a biologically inspired framework well-suited to these challenges due to their natural ability to process discrete events and preserve millisecond-scale timing. This work investigates the application of SNNs to high-frequency price-spike forecasting, enhancing performance via robust hyperparameter tuning with Bayesian Optimization (BO). This work converts high-frequency stock data into spike trains and evaluates three architectures: an established unsupervised STDP-trained SNN, a novel SNN with explicit inhibitory competition, and a supervised backpropagation network. BO was driven by a novel objective, Penalized Spike Accuracy (PSA), designed to ensure a network's predicted price spike rate aligns with the empirical rate of price events. Simulated trading demonstrated that models optimized with PSA consistently outperformed their Spike Accuracy (SA)-tuned counterparts and baselines. Specifically, the extended SNN model with PSA achieved the highest cumulative return (76.8%) in simple backtesting, significantly surpassing the supervised alternative (42.54% return). These results validate the potential of spiking networks, when robustly tuned with task-specific objectives, for effective price spike forecasting in HFT.

Paper Structure

This paper contains 23 sections, 12 equations, 5 figures, 8 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Method
  3. Data and Preprocessing
  4. Volume Weighted Average Prices
  5. Feature Creation
  6. Normalisation
  7. Encoding
  8. Spike Definitions and Predictive Metrics
  9. Real vs Fake Price Spikes
  10. Momentum vs Reversion Spikes
  11. Predictive Metrics
  12. The Penalised Spike Accuracy (PSA) Objective
  13. Model Architectures
  14. Experimental Design
  15. SNN Based Momentum Strategy
  16. ...and 8 more sections

Figures (5)

  • Figure 1: Comparison of the raw transaction price (left) and the VWAP (right).
  • Figure 2: Comparison of trend components in price vs. differenced price series using unobserved component analysis (UCA) durbin2012timeseriesseabold2010statsmodels.
  • Figure 3: The extended architecture incorporating multiple time lags (additional nodes in $X_1 \text{ and }X_2$) and explicit inhibitory connections (in red).
  • Figure 4: Parameter importance ranking for Models 1 & 2, explored for performance metrics SA and PSA.
  • Figure 5: Equity (unscaled) and Drawdown over time.