EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
3520. A methodology of predict-interpret-suggest for used car pricing
Invited abstract in session TD-6: Advancements of OR-analytics in statistics, machine learning and data science 15, stream Advancements of OR-analytics in statistics, machine learning and data science.
Tuesday, 14:30-16:00Room: 1013 (building: 202)
Authors (first author is the speaker)
| 1. | Qihao Wu
|
| Department of of Data and Systems Engineering, The University of Hong Kong | |
| 2. | Yong-Hong Kuo
|
| Department of of Data and Systems Engineering, The University of Hong Kong |
Abstract
Pricing used cars is a complicated process that involves various influencing factors. This study deploys a series of interpretable machine learning models to predict used car prices with elaborate feature engineering. A comparative analysis reveals that ensemble tree algorithms can achieve the highest performance, with an R-squared value exceeding 0.93. Through examining the interpretable feature attributes, the used cars' age, make, and engine capacity are ranked as the top three most significant determinants of price. Furthermore, our wrapper method explores the heterogeneity of features and datasets, by which the performance of neural networks with numerical features on more specific data is highlighted. Finally, the treatment effects of significant binary variables are summarized as suggestions for consumers. By integrating predictive modeling, interpretability, and causal inference, this study presents a novel Predict-Interpret-Suggest framework for the used car market.
Keywords
- Analytics and Data Science
- Machine Learning
- Forecasting
Status: accepted
Back to the list of papers