How to Build and Predict Clustered Industry Index Return on Continuous Time Series Data with Non-Continuous Values

How to Build and Predict Clustered Industry Index Return on Continuous Time Series Data with Non-Continuous Values

• Tian Tian, Stuart Management Science Ph.D. student

Abstract:

Normal methods of time series analysis perform poorly for stock selection. This is largely because these series do not conform to the continuously evolving process assumptions inherent in their specification. The purpose of this research is to use advanced machine learning techniques to fix this shortcoming. With this technique we can predict the time series of an index of correlated stocks very accurately. We build the index using Hierarchical Risk Parity (HRP) (Lopez De Prado 2016) and after documenting the noisy, chaotic nature of the series we forecast returns using CNN-LSTM and LSTM-CNN based GAN modeling. This form of modeling performs better than the popular LSTM machine learning algorithm.

All Illinois Tech faculty, students, and staff are invited to attend.

The Friday Research Presentations series showcases ongoing academic research projects conducted by Stuart School of Business faculty and students, as well as guest presentations by Illinois Tech colleagues, business professionals, and faculty from other leading business schools.