基于改进粒子群算法的LSTM混合神经网络期权定价模型  

Option Pricing Model of LSTM Hybrid Neural Network Based on Improved Particle Swarm Optimization

作　　者：章伟果[1] 龚武胜 扈文秀[1] ZHANG Wei-guo;GONG Wu-sheng;HU Wen-xiu(School of Economic and Management,Xi'an University of Technology,Xi'an 710054,China)

机构地区：[1]西安理工大学经济与管理学院,陕西西安710054

出　　处：《系统工程》2024年第1期139-148,共10页Systems Engineering

基　　金：国家自然科学基金项目(71603203,71971169);陕西省软科学项目。

摘　　要：引入改进粒子群算法(IPSO)对长短时记忆神经网络模型(LSTM)超参数进行自适应匹配,并结合Heston模型进行混合建模,提出一种全新的IPSO-LSTM-Heston期权定价模型。为验证模型的定价效果,基于上证50ETF期权高频数据进行实证分析。结果表明:IPSO算法具有优异的全局寻优能力和收敛速度,能够大幅提高LSTM混合神经网络模型的定价效率。通过将优化的LSTM神经网络模型与Heston模型结合,不仅可以捕捉高频数据的动态特征,而且能够发挥神经网络模型的非线性拟合能力与传统模型定价过程的严密性等优点,在降低定价误差的同时显著提高定价精准度。This paper introduces IPSO algorithm to adaptively match the hyperparameter of LSTM neural network,and proposes a new IPSO-LSTM-Heston option pricing model combined with Heston model.In order to test the pricing effect,an empirical analysis is carried out based on the high-frequency data of 50ETF option.The results show that:IPSO algorithm has excellent global optimization ability and convergence speed,which can greatly improve the pricing efficiency of LSTM hybrid neural network model.By combining the LSTM neural network model optimized by IPSO algorithm with Heston model,the model proposed in this paper can not only capture the dynamic characteristics of high-frequency data,but also give full play to the nonlinear fitting ability of neural network model and the tightness of traditional model pricing process,so as to significantly improve the pricing accuracy while reducing the pricing error of the model.

关 键 词：期权定价 粒子群算法 LSTM神经网络 混合神经网络模型 

分 类 号：F830[经济管理—金融学]