基于PSO-HKELM的内部交易识别
Research on insider trading identification based on PSO-HKELM
邢蕾 1刘艳彩1
作者信息
- 1. 长春工业大学 数学与统计学院,吉林 长春 130012
- 折叠
摘要
对于证券市场出现的内部交易问题,收集了 2018-2022 年中国证监会公布的对内部交易惩罚的公司股票数据作为样本,从我国证券市场表现、财务表现、股权结构以及治理体系三个方面选择相关指标,提出一种粒子群优化混合核极限学习机(HKELM)的算法,建立相应的内部交易行为识别模型.实验结果表明,提出的PSO-HKELM 模型效果较好,平均准确率为 79.68%,比 HKELM、ELM、RF 分别高 4.27%、6.32%、11.22%,可以看出,粒子群对HKELM进行了有效优化,提高了识别效率,在时间窗口期为 90d 时结果最佳且稳定.有助于监管部门准确把握发生的内部交易行为,进一步提高内部交易识别能力.
Abstract
For the insider trading problems in the securities market,this paper collects the stock data of companies punishing for insider trading published by the China Securities Regulatory Commission from 2018 to 2022 as a sample,selects relevant indicators from three aspects:China's securities market performance,financial performance,equity structure and governance system,proposes an algorithm for particle swarm optimization HKELM,and establishes a corresponding insider trading behavior recognition model.The experimental results show that the PSO-HKELM model proposed in this paper has a good effect,with an average accuracy of 79.68%,which is 4.27%,6.32%and 11.22%higher than HKELM,ELM and RF.Results were optimal and stable with a time window of 90 days.It helps the regulatory authorities accurately grasp the insider transactions that occur and further improves the ability to identify insider transactions.
关键词
内部交易/粒子群/混合核极限学习机/行为识别Key words
insider trading/particle swarm optimization/HKELM(Hybrid Kernel Extreme Learning Machine)/behavioral recognition引用本文复制引用
基金项目
吉林省教育厅科学研究项目(JJKH20230743KJ)
出版年
2024
NETL
NSTL
万方数据