A Bayesian Investment Analyst on the Johannesburg Stock Exchange

Wilson Tsakane Mongwe (), Rendani Mbuvha () and Tshilidzi Marwala
Additional contact information
Wilson Tsakane Mongwe: University of Johannesburg
Rendani Mbuvha: University of Witwatersrand
Tshilidzi Marwala: United Nations University

Chapter Chapter 13 in Bayesian Machine Learning in Quantitative Finance, 2025, pp 281-311 from Springer

Abstract: Abstract Sell-side analyst forecasts are relied upon by investors for their portfolio and risk management decisions. Evaluating such reports’ accuracy is thus important for investors and other stakeholders. In Sidogi et al. (2022), we explored the use of traditional machine learning techniques to model the bidirectional accuracy of sell-side analyst forecasts on the Johannesburg Stock Exchange. This chapter extends this work by framing it within a Bayesian framework and thus introducing a “Bayesian Investment Analyst.” We train a Bayesian logistic regression (BLR) model using the Metropolis-Adjusted Langevin Algorithm, Hamiltonian Monte Carlo, and Separable Shadow Hamiltonian Hybrid Monte Carlo Markov Chain Monte Carlo (MCMC) techniques. We utilize an automatic relevance determination (ARD) prior distribution for our BLR model to create the BLR-ARD model. The BLR-ARD model allows us to automatically rank the most relevant features for determining the bidirectional accuracy of analyst forecasts. We consider four experiments: (1) using the full dataset, (2) a dataset focusing only on resource stocks, (3) a dataset related only to financial stocks, and 4) a dataset only focusing on industrial stocks. Our results show that, across all the MCMC methods, the BLR-ARD model has the best performance on the industrial stocks followed by the resources stocks, with the worst performance being on the financial stocks. We further find that the MCMC methods agree on which features are the most relevant for this task. The essential features vary across the different sectors, with relative volatility and initial price being most important for the entire dataset; rolling 20day volatility and target price being critical on the resource stocks; analyst coverage, initial price, and target price being important for financial stocks while the target price, initial price, analyst coverage, and Top40 index cross-sectional volatility being significant for industrial stocks.

Date: 2025
References: Add references at CitEc
Citations:

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:sprchp:978-3-031-88431-3_13

Ordering information: This item can be ordered from
http://www.springer.com/9783031884313

DOI: 10.1007/978-3-031-88431-3_13

Access Statistics for this chapter

More chapters in Springer Books from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().