University of Twente Student Theses

Login

Temporal Aspects of Stock Price Prediction : Quantifying the Role of Historical Data using Partitioned Dynamic Bayesian Networks

Verdecchia, Cristian (2024) Temporal Aspects of Stock Price Prediction : Quantifying the Role of Historical Data using Partitioned Dynamic Bayesian Networks.

[img] PDF
3MB
Abstract:Despite the abundance of studies in the financial markets area, few have investigated Bayesian Networks for stock price predictions, which are traditionally perceived as unsuitable for financial forecasting due to their inherent complexities. Despite this, we believe that the S&P 500 index, known to be a good reflection of the status of the American economy, presents an ideal case study for this research. Comprising a vast amount of companies across various sectors, the S&P 500 offers a rich dataset for analysis. Despite their complexities, the explainability that BNs inherently bring, could in the future, provide important information on hidden market dynamics. Leveraging PDBNs, originally designed for health-related data, provides a unique opportunity to test the evolving behavior of the market. Their ability to change in structure provides a great advantage over the more traditional DBNs when dealing with this type of data. In this thesis, great emphasis was dedicated to acquiring high-quality data tailored for this approach. Multiple networks were created capturing various aspects of PDBNs through which the effects of historical data on predictions were analysed. The models were compared to evaluate their performances where the final employed model provided a good balance between specificity and sensitivity.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science MSc (60300)
Link to this item:https://purl.utwente.nl/essays/103860
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page