Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Multistep Prediction of Multivariate Market Indices Using Weighted Optical Reservoir Computing

Fang Wang, Ting Bu, Yuping Huang

TL;DR

This paper introduces a weighted free-space optical reservoir computing approach to multivariate stock index forecasting, integrating seven indices with seven macroeconomic/technical features. By encoding temporal data into high-dimensional optical signals and applying feature-correlation weights, the method achieves superior one-step and multi-step predictions compared with traditional ML and LSTM models, even with limited training data. The results demonstrate that the optical RC framework can better capture market volatility and nonlinear dynamics, offering a low-energy, real-time, parallel processing solution for financial forecasting. The work suggests significant practical impact for edge pricing and rapid decision-making in volatile markets, with future directions focusing on transitioning more processing steps into the optical domain.

Abstract

We propose and experimentally demonstrate an innovative stock index prediction method using a weighted optical reservoir computing system. We construct fundamental market data combined with macroeconomic data and technical indicators to capture the broader behavior of the stock market. Our approach shows significant higher performance than state-of-the-art methods such as linear regression, decision trees, and neural network architectures including long short-term memory. It captures well the market's high volatility and nonlinear behaviors despite limited data, demonstrating great potential for real-time, parallel, multi-dimensional data processing and predictions.

Enhancing Multistep Prediction of Multivariate Market Indices Using Weighted Optical Reservoir Computing

TL;DR

This paper introduces a weighted free-space optical reservoir computing approach to multivariate stock index forecasting, integrating seven indices with seven macroeconomic/technical features. By encoding temporal data into high-dimensional optical signals and applying feature-correlation weights, the method achieves superior one-step and multi-step predictions compared with traditional ML and LSTM models, even with limited training data. The results demonstrate that the optical RC framework can better capture market volatility and nonlinear dynamics, offering a low-energy, real-time, parallel processing solution for financial forecasting. The work suggests significant practical impact for edge pricing and rapid decision-making in volatile markets, with future directions focusing on transitioning more processing steps into the optical domain.

Abstract

We propose and experimentally demonstrate an innovative stock index prediction method using a weighted optical reservoir computing system. We construct fundamental market data combined with macroeconomic data and technical indicators to capture the broader behavior of the stock market. Our approach shows significant higher performance than state-of-the-art methods such as linear regression, decision trees, and neural network architectures including long short-term memory. It captures well the market's high volatility and nonlinear behaviors despite limited data, demonstrating great potential for real-time, parallel, multi-dimensional data processing and predictions.
Paper Structure (8 sections, 4 equations, 7 figures, 2 tables)

This paper contains 8 sections, 4 equations, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Model
  3. Experiment diagram
  4. Datasets preparation
  5. Optical RC
  6. Multiple ML and LSTM models
  7. Results and analysis
  8. Conclusion

Figures (7)

  • Figure 1: Experiment diagram for stock benchmark prediction to compare spatial reservoir computing and competitive machine learning models.
  • Figure 2: (a) An overview of one stock index sequence $^$NYA with training and testing parts. (b) A feature correlation matrix heatmap for $^$NYA
  • Figure 3: The configuration of the spatial reservoir computer. SLM: spatial light modulator, BS: beam splitter, HWP: half-wave plate, QWP: quarter-wave plate, PBS: polarizing beam splitter, PM: power meter
  • Figure 4: The mapping method to combine multi-feature inputs and the camera state for spatial light modulation.
  • Figure 5: True values versus 1-step prediction results from optical RC with and without feature correlation weights, contrast to the best-performing ML results, shown for $^$NYA (a) and $^$N225 (b).
  • ...and 2 more figures