COGSPMS
An Intelligent System for Automated Share Evaluation using Graph Neural Networks and Semantic Bayesian Networks
Abstract
Portfolio management is a decision-making process that aims to minimise risk and maximise return through the proportional allocation of capital into identified financial securities or shares. The initial share evaluation phase involves identifying shares with suitable risk-return characteristics for inclusion in an investment portfolio.
Semantic Bayesian networks are a class of Artificial Intelligence techniques that support explainability in intelligent systems. This research implements and evaluates the INVEST intelligent decision support system and several extensions under various conditions for Johannesburg Stock Exchange-listed share evaluation.
Spatial-temporal graph neural networks are models that process graphically-encoded multivariate data exhibiting both spatial and temporal dependencies. This research also evaluates ST-GNN architectures for Johannesburg Stock Exchange price prediction, and assesses the suitability of a correlation matrix to capture market dependencies and encode structural information.
Videos 1
Watch presentations, demos, and related content
COGSPMS: An Intelligent System for Automated Share Evaluation using Graph Neural Networks and Semantic Bayesian Networks
Like, comment, and subscribe on YouTube to support the creator!
COGSPMS: An Intelligent System for Automated Share Evaluation using Graph Neural Networks and Semantic Bayesian Networks
Documents 4
Downloadable resources and documentation
Gallery 6
Explore the visual story of this exhibit
Value Evaluation Network
Quality Evaluation Network
Investment Recommendation Network
INVEST Console
JSE Top 40 Index Correlation Matrix
Graph WaveNet Adaptive Graph