2862. Exact and Heuristic Solution Approaches for Clustering with Localized Feature Selection
Invited abstract in session MD-34: Advancements of OR-analytics in statistics, machine learning and data science 1, stream Advancements of OR-analytics in statistics, machine learning and data science.
Monday, 14:30-16:00Room: Michael Sadler LG10
Authors (first author is the speaker)
| 1. | Cem Iyigün
|
| Department of Industrial Engineering, Middle East Technical University (METU) | |
| 2. | Sinan Gürel
|
| Industrial Engineering Department, Middle East Technical University | |
| 3. | Gozdenur Buyuk
|
| Industrial Engineering, Middle East Technical University |
Abstract
The complexity of real-world data characterized by excessive number of features
poses significant challenges in clustering problems. Incorporating feature
selection decisions in clustering is a way to tackle this issue. Clustering
with localized feature selection takes into account the fact that, for each cluster,
the relevant set of features may be different. Localized feature selection
has been in use in many areas of clustering, such as image segmentation,
bioinformatics, and recommendation systems. This paper studies a clustering
problem that makes joint clustering and localized feature selection decisions
The paper presents alternative Second Order Cone Programming models that find exact optimum solutions. Furthermore, it presents a mathematical programming-based heuristic
(matheuristics) and a heuristic algorithm for the problem. Finally, the paper
compares computational performance of proposed methods.
Keywords
- Machine Learning
- Programming, Mixed-Integer
- Algorithms
Status: accepted
Back to the list of papers