1461. A SHAP-Driven Approach for Explainable Ensemble Pruning Using Semi-Infinite Programming
Invited abstract in session MB-4: Interpretable Optimization Methods and Applications, stream Data Science meets Optimization.
Monday, 10:30-12:00Room: Rupert Beckett LT
Authors (first author is the speaker)
| 1. | Melisa Caliskan-Demir
|
| Industrial Engineering, Istanbul Aydin University | |
| 2. | Volkan Bakır
|
| Graduate School Department of Artificial Intelligence, Bahcesehir University | |
| 3. | Sureyya Ozogur-Akyuz
|
| Department of Mathematics Engineering, Bahcesehir University |
Abstract
In this study, an ensemble Convolutional Neural Network (CNN) model was developed using different parameter configurations on the different datasets. The ensemble was pruned using Semi-Infinite Programming (SIP) to reduce model complexity while maintaining classification performance. Following this, SHapley Additive exPlanations (SHAP), an Explainable Artificial Intelligence (XAI) technique, was incorporated to enhance the interpretability of the model. SHAP values were used to analyze the contribution of different features to the decision-making process, and a second ensemble was formed based on this explainability-driven analysis. This new ensemble was then pruned again using SIP, and the results were compared with the initial pruning approach. Experimental findings demonstrate that SIP-based pruning effectively optimizes ensemble performance, while integrating SHAP enhances transparency by identifying key image regions influencing predictions. The comparative analysis of pruned ensembles, with and without SHAP integration, highlights the impact of explainability-driven pruning on deep learning model efficiency and decision reliability.
Keywords
- Artificial Intelligence
- Machine Learning
- Optimization Modeling
Status: accepted
Back to the list of papers