Abstract—In contrast to their outstanding success in dealing with unstructured data, such as images and natural language, machine-learning models have not shown noticeable achievements in utilizing structured data, i.e., tabular-format data. Part of their excellent performance with unstructured data comes from their capability of transfer learning, which has rarely been utilized in the fields of structured data. In this study, a random forest is used to estimate vehicle prices in the South Korean automobile industry. To enhance the performance of the random forest, when the input variables are structured data and part of them are high-cardinality categorical types, entity embedding vectors are created from a neural network, and reused in the random forest training. This study demonstrates that information in structured data can be efficiently extracted using the entity embedding technique and effectively reused in different but related tasks in the form of transfer learning.
 

Index Terms—machine learning, transfer learning, structured data, entity embedding, random forest

Cite: Changro Lee, "Random Forest with Transfer Learning: An Application to Vehicle Valuation," Journal of Advances in Information Technology, Vol. 13, No. 4, pp. 326-331, August 2022.

Copyright © 2022 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.