Generalized PCR Model for Stock Price Modelling

Course RMSC4002 – Financial Data Analytics with Machine Learning
Completed on
Project type Group project

This group project develops a generalized principal component regression approach for stock price modelling and stock selection. The focus was on handling high-dimensional financial factors through PCA-based compression before prediction.

Highlights

  • Implemented a generalized PCR pipeline in R.
  • Worked with high-dimensional equity factor data sourced from Bloomberg.
  • Compared PCA-based modelling with more standard regression baselines.

Methods

  • Used PCA to reduce dimensionality in factor-ratio predictors.
  • Applied the compressed factor representation in a regression framework for return estimation.
  • Evaluated model performance against simpler benchmark approaches.

Findings

  • The project provided a practical example of how dimensionality reduction can improve financial modelling workflows.
  • It also highlighted the tradeoff between interpretability and predictive efficiency in factor-based models.

Resources