Page 320 - Data Science Algorithms in a Week

In: Artificial Intelligence ISBN: 978-1-53612-677-8
Editors: L. Rabelo, S. Bhide and E. Gutierrez © 2018 Nova Science Publishers, Inc.

Chapter 13

PREDICTIVE ANALYTICS FOR THERMAL

COAL PRICES USING NEURAL NETWORKS

AND REGRESSION TREES

2,
1
Mayra Bornacelli , Edgar Gutierrez and John Pastrana 3
*
1 Carbones del Cerrejón (BHP Billiton, Anglo American, Xtrata), Bogota, Colombia
2 Center for Latin-American Logistics Innovation, Bogota, Colombia
3 American Technologika, Clermont, Florida, US

ABSTRACT

The research is aimed at delivering predictive analytics models which build powerful
means to predict thermal coal prices. The developed methodology started by analyzing
expert market insights in order to obtain the main variables. The Delphi methodology was
implement in order to reach conclusions about the variables and tendencies in the global
market. Then, artificial intelligence techniques such as neural networks and regression
trees were used to develop and refine the number of variables. The predictive models
created were validated and tested. Neural networks outperformed regression trees.
However, regression trees created models which were easy to visualize and understand.
The conceptual results from this research can be used as an analytical framework to
facilitate the analysis of price behavior (oligopolistic markets) to build global business
strategies.

Keywords: predictive analytics, neural networks, regression trees, thermal coal price

* Corresponding Author Email: edgargutierrezfranco@gmail.com