.png){kind=icon}
.png){kind=icon}
Return
Graph-based edge betweenness centrality in multi-view ensemble frameworks
.png){kind=icon}
PRE
AI
DOI：10.1007/s12065-025-01108-9.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The view construction (subsets of feature set) for the single-source dataset is a challenging job in multi-view learning to achieve better performance than traditional machine learning strategies. The two factors that have impacted the effectiveness of multi-view learning hugely are the quantity of views and the quality of each generated view. In this research, the edge betweenness centrality-based view construction (EBC-VC) method has been proposed, specifically for low-dimensional datasets. It finds the number of views based on edge betweenness and ensures the quality of individual views based on correlation parameters. Eight standard datasets have been used to examine the validity of the proposed method. The k-nearest Neighbours (KNN), Support Vector Machine (SVM), Na & iuml;ve Bayes (NB), Neural Network (NN), and Decision Tree (DT) classifiers have been deployed to learn over each view. A comparative study based on the classification accuracy assessment parameter and its non-parametric statistical analysis using the Friedman ranking has been conducted to illustrate the potential of the suggested framework. The proposed EBC-VC method outperformed single-view learning and other feature set partitioning techniques, according to the experimental data and their statistical analysis.
Keywords:
Views construction
Edge betweenness centrality
Multi-view ensemble learning
Minimum spanning tree
Feature set partitioning
Journal
IF:
2.6
Papers:
116
Citations:
2.0K