What is it about?

Feature Selection plays an important role in modelling data with high dimensions (or many variables), which happens to be the case in most real world applications. The existing techniques in general adopts a data-oriented approach, in a way that the selection is solely based on quantitative evaluation, ignoring the underlying context of the data. This paper proposed a new product-driven approach by incorporating domain-knowledge from industrial experts to support the selection process, which has shown great effectiveness in the experiment of predicting sailboat prices in real transaction records.

Featured Image

Why is it important?

The Domain-Knowledge based Feature Selection Framework (DKFS) proposed in this research stands out for its enhanced robustness and effectiveness in dimensionality reduction. This reduction contributes to improved model interpretability, decreased computational complexity, and the potential for enhanced generalization to new data. While our framework is tested under the sailboat price prediction, it is expressly tailored for product-oriented applications in general with great adaptability across diverse domains. This versatility enhances its practical utility, making it applicable to a broad spectrum of industries. Last but not least, the integration of domain knowledge in traditional statistical methods is innovative and motivates further research in this promising field, particularly the research of expert systems and adaptive feature selection design with their real world applications.

Perspectives

I believe the results and discussions presented in this paper are pretty much derived from the general human-intuition, as domain knowledge itself comes from people in general. This is a great attempt in combining theoretical research in feature selection and the real world business applications, which turned out to be a successful one with great potential in making a huge impact in the future.

MengZe Hong
University of Nottingham

Read the Original

This page is a summary of: Novel Domain-Knowledge Based Feature Selection Framework for Price Prediction: Comprehensive Modelling in Sailboat Market, September 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3629264.3629284.
You can read the full text:

Read

Contributors

The following have contributed to this page