Volume 22, Issue 2 (9-2025)
JSDP 2025, 22(2): 127-138 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Nasiri M, Daneshpour N. Presenting a new method for multi label classification based on neural network. JSDP 2025; 22 (2) : 8
URL: http://jsdp.rcisp.ac.ir/article-1-1433-en.html
URL: http://jsdp.rcisp.ac.ir/article-1-1433-en.html
Shahid Rajaee Teacher Training University
Abstract: (31 Views)
The problem of classification can be divided into two categories: single-label classification and multi-label classification. Binary classification and multi-class classification are two types of single-label classification. Binary classification refers to predicting one of two classes and multi-class classification involves predicting one of more than two classes. In contrast to traditional supervised learning, in multi-label learning, each object is represented by set of labels instead of a single label. Samples in Multi-label classification can have zero, one or more than one label. In other words, each instance is represented by a set of labels. Some approaches like Binary Relevance (BR), Transforms the problem into multiple independent binary classification tasks, one for each label. While simple, it ignores label correlations; some other algorithms like Label Powerset (LP) have shown that considering the relationship between labels can help to improve the results. In this paper, to consider the correlation between labels and also to solve the imbalanced classification problem, k-means constraint clustering is used for labels and features in the first step. In the second step, a separate multi layers neural network is considered for each cluster of labels. Due to the use of separate classifications and the increase in training time, a new method has been used to reduce dimensions using scatter add. Attributes are sent to the input of each neural network using the scatter add method, and the output of each neural network is the labels in that desired cluster. Finally, the final labels are obtained by combining the labels predicted by the classifiers. By evaluating the proposed method on the existing data set in comparison with the existing methods, we come to the conclusion that in many criteria such as accuracy, precision and hamming-loss among the three text data sets, it has the first rank with a difference of one percent.
Article number: 8
Type of Study: Research |
Subject:
Paper
Received: 2024/07/14 | Accepted: 2025/03/15 | Published: 2025/09/13 | ePublished: 2025/09/13
Received: 2024/07/14 | Accepted: 2025/03/15 | Published: 2025/09/13 | ePublished: 2025/09/13
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
