Science Belongs to All Mankind

JOURNALS-DETAIL

深度学习的互联网金融投诉文本分类研究

Author:

邢智毅

Vol. 2, Issue 3, Pages: 44-46(2025)

Doi:

https://doi.org/10.62639/sspis13.20250203

ISSN:

3006-0737

EISSN:

3006-4309

Views:

27

Downloads:

0

Abstract References Project Publication Info Metrics
Abstract

本文提出一种基于深度学习的分类算法。该方法结合多标签注意力机制和图神经网络,精细提取文本特征并建模标签间关系,提升了尾部低频标签的识别能力。同时,通过非对称损失函数优化模型,缓解正负样本不均衡问题,显著提高了少量投诉类别的召回率。实验结果表明,本文方法在所有标签和尾部标签的宏观F1值上分别较传统方法提升3.1%和6.9%,在互联网金融投诉分类中具有较高的应用价值。

Keyword

互联网金融;投诉文本;多标签分类;深度学习;图神经网络

International Scientific Studies Press Limited

International Scientific Studies Press Limited is a company boasting rich international communication resources and formidable editorial, translation, and publishing capabilities. Our primary focus revolves around the publication of academic journals. Our establishment's mission is to provide a premier publishing platfor...
FLAT C,23/F,LUCHY PLAZA,315-321 LOCKHART ROAD,WANCHAI,HONG KONG (00852) 65557188
Copyright © 2025 International Scientific Studies Press Limited