Learning Non-Linear Relationships in the Cross-Section

Asset managers use factor investing to explain and forecast individual asset returns and to construct portfolios. A factor approach can help to increase diversification, improve returns, and minimize risks. Classic factors include value, size, and momentum, but the list of modern “candidate” factors is exceedingly long.

Credit Suisse’s equity investment strategist Ricardo Pachón Cortes and Valerio Sperandeo of MathWorks® discuss how to effectively apply machine learning techniques to a range of factors available in the HOLT dataset.

In this presentation, they explain:

- How to apply factor-timing techniques as well as machine learning techniques to choose between factor weights

- How to optimize the hyperparameters in the fitting process

- The outputs of the machine learning process

     Introduction

     Why use MATLAB?

     What is Equity Factor Investing

     Factor Prediction

     Machine Learning Model for Stock Selection

     Training Data

    HOLT Variable and Equity Factor Definitions

    Hyperparameter Optimization

   The Results

    Interpreting the Results

    Putting it into Practice