EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
1771. Outranking hypercuboid learning approach for classification problems
Invited abstract in session WA-28: Advancements of OR-analytics in statistics, machine learning and data science 8, stream Advancements of OR-analytics in statistics, machine learning and data science.
Wednesday, 8:30-10:00Room: 065 (building: 208)
Authors (first author is the speaker)
| 1. | Nabil Belacel
|
| Data Science for Complex Systems |, Digital Technologies Research CenterNational Research Council-ICT |
Abstract
Classifiers in today's data-driven domain face myriad challenges: overfitting, computational overheads, diminished accuracy, imbalanced datasets, and the opaque "black box" issue. Conventional classifiers struggle with managing noisy and missing features. This study introduces classification methods leveraging hypercuboid learning and outranking measures.
Our framework autonomously discerns hypercuboid features from the training set, eliminating the need for prior domain expertise. These hypercuboids capture essential dataset patterns, while outranking measures mitigate noise and uncertainty. Empirical evaluations across diverse datasets from the UCI repository compare the proposed classifiers with established models like k-NN, SVM, Random Forest, Neural Networks, and Naive Bayes. Results demonstrate our classifiers' robustness against imbalanced data and extraneous features, achieving comparable or superior performance to benchmarks. Moreover, our models offer interpretability without sacrificing predictive accuracy.
Keywords
- Machine Learning
- Expert Systems and Neural Networks
- Computer Science/Applications
Status: accepted
Back to the list of papers