Paper: arXiv 2307.08665

Authors: Nelson Kyakutwika, Bruce Bartlett

Abstract

Cross-series dependencies are crucial in obtaining accurate forecasts when forecasting a multivariate time series. Simultaneous Graphical Dynamic Linear Models (SGDLMs) are Bayesian models that elegantly capture cross-series dependencies. This study forecasts returns of a 40-dimensional time series of stock data from the Johannesburg Stock Exchange (JSE) using SGDLMs. The SGDLM approach involves constructing a customised dynamic linear model (DLM) for each univariate time series. At each time point, the DLMs are recoupled using importance sampling and decoupled using mean-field variational Bayes. Our results suggest that SGDLMs forecast stock data on the JSE accurately and respond to market gyrations effectively.

Complexity vs Empirical Score

  • Math Complexity: 9.0/10
  • Empirical Rigor: 5.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced Bayesian state-space modeling, variational inference, and importance sampling, demanding high mathematical sophistication. It applies the model to 40 stocks from the Johannesburg Stock Exchange with CPU-based implementation, indicating solid empirical validation, though the hardware specifications are modest compared to GPU-accelerated industry standards.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Forecast JSE Stock Returns<br>using SGDLMs"] --> B["Data Input<br>40-dimensional time series<br>from Johannesburg Stock Exchange"]
  B --> C["Model Construction<br>Customised DLM per<br>univariate series"]
  C --> D{"At each time point"}
  D --> E["Recoupling via<br>Importance Sampling"]
  D --> F["Decoupling via<br>Mean-field Variational Bayes"]
  E & F --> G["Computational Process<br>Bayesian inference<br>capturing cross-series dependencies"]
  G --> H["Key Findings<br>Accurate forecasts<br>Effective response to market gyrations"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes using Simultaneous Graphical Dynamic Linear Models to forecast daily returns of 40 JSE stocks.

Source sentence
This study forecasts returns of a 40-dimensional time series of stock data from the Johannesburg Stock Exchange (JSE) using SGDLMs.

Assumes 5

  • All stocks have the same number k of simultaneous parents. model
    Source sentence
    In this paper, we follow Gruber and West 2016 and stipulate that all stocks have the same number 𝑘 of simultaneous parents, i.e., |sp(𝑖)| = 𝑘 for all 𝑖
  • The initial prior covariance matrix is block‑diagonal and this structure is preserved over time. model
    Source sentence
    Given that the matrix 𝐑0 in the initial prior is typically block-diagonal, we strive to preserve this block-diagonal structure for 𝐑𝑡 at all times 𝑡.
  • The data consist of daily log‑returns of the selected stocks. data
    Source sentence
    Our data set is the daily log-returns of 40 JSE stocks that were selected from the Top 100 JSE index.
  • CPU‑based computers are sufficient for a low‑dimensional (40‑stock) SGDLM analysis. execution
    Source sentence
    using the common central processing unit (CPU)-based computers should suffice when one has a smaller number of time series, for example, the 40 stocks of the current study.
Show 1 more
  • Simultaneous parents are selected once at observation 782 and kept fixed for the rest of the analysis. execution
    Source sentence
    In the current study, we picked the simultaneous parent of each of the stocks at 𝑡= 782 and maintained it to the end of the analysis.

Data

  • Daily log‑returns of 40 JSE stocks · Yahoo! Finance market-prices public
    01/01/2014 to 30/06/2022 · Top 100 JSE index constituents
    Source sentence
    ... We downloaded the daily closing prices of all the stocks from Yahoo! Finance for the period 01/01/2014 to 30/06/2022 and calculated the log-returns.

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • SGDLM forecasts the returns of the stock data accurately. predictive-accuracy
    Source sentence
    Our study has shown that the SGDLM forecasts the returns of the stock data accurately.
  • With 40 stocks or less, the most accurate forecasts are obtained with one simultaneous parent. predictive-accuracy
    Source sentence
    With a dimension of 40 stocks or less, our results suggest that the most accurate forecasts are obtained with one simultaneous parent.
  • SGDLM outperforms DLM for Standard Bank, British American Tobacco and Shoprite Holdings Limited (both RMSE and MAD smaller). performance
    Source sentence
    The SGDLM performs better than the DLM in the case of Standard Bank, British American Tobacco and Shoprite Holdings Limited given that it gives smaller values of both the RMSE and the MAD
  • For Glencore plc, Anglo American plc and MTN Group Limited, DLM yields smaller errors than SGDLM. performance
    Source sentence
    the exact opposite occurs for Glencore plc, Anglo American plc and MTN Group Limited.
  • Overall differences between SGDLM and DLM errors are very small. statistical
    Source sentence
    It can be seen that the differences between the errors of the SGDLM and the DLM are very small.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 27 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 1 paper in the archive (1 by independent authors)

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.