Predicting Industries and Productivity

Authors

• Taylor Kaye, St. John Fisher University
• Kyler Patrick, St. John Fisher University

Document Type

Poster Presentation

Publication Date

4-17-2026

Keywords

fsc2026

Abstract

This project uses total factor productivity data from the U.S. Bureau of Labor Statistics to model and predict both industry classification and productivity growth. Using a dataset of 2,911 observations with numerous economic inputs, we applied tree-based machine learning methods, specifically conditional inference trees, ctrees, and random forest models. After preprocessing steps such as handling missing data, scaling, and train-test splitting, models were evaluated on their predictive performance. Results show that random forests outperformed ctrees demonstrating strong capability in distinguishing among 81 industries. Key predictors of productivity growth included capital intensity, labor input, and energy-related indices. Overall, the findings highlight the effectiveness of tree-based methods in capturing complex economic relationships and suggest that industry structure is strongly driven by input composition and productivity dynamics.

Comments

Poster presented at the 2026 Fisher Showcase, St. John Fisher University, April 17, 2026.

Publication Information

Kaye, Taylor and Patrick, Kyler, "Predicting Industries and Productivity" (2026). Fisher Showcase 2026. Paper 153.
https://fisherpub.sjf.edu/fsc2026/153

Please note that the Publication Information provides general citation information and may not be appropriate for your discipline. To receive help in creating a citation based on your discipline, please visit https://libguides.sjf.edu/citations.