摘要
对金融衍生品的定价一直都是理论界和实务界所关注的热点问题之一。传统的参数模型定价方法,容易产生系统性偏差,而神经网络方法在解决这一问题时具有明显的优势。本文基于遗传算法改进的BP神经网络模型(GABP),利用我国8只权证的1050天数据、标的股票1900天的数据和其他所需数据作为训练数据训练网络。研究结果表明,神经网络在权证的定价中效果要优于B-S模型;而RBF模型精度大于BP模型的预测,GABP模型的精度大于RBF模型的预测。
How to price financial derivates is one of hot topics for theorists and practice fields. The idea of the traditional parameter pricing method can easily lead to systematic bias, but neural network has the obvious superiority. In this article we use the genetic algorithms to improve the BP neural network model （counted as GABP model）, analyze the China＇ s eight warrants data of 1050 days, the underlying stock data and other data. At last we found that a neural network in the pricing of warrants in effect is better than the B-S model in effect. The RBF neural network is better prediction in accuracy than the BP neural network, the accuracy of the GABP model larger than that of the RBF neural network.
出处
《金融理论与实践》
北大核心
2010年第9期22-27,共6页
Financial Theory and Practice
关键词
BP神经网络
RBF神经网络
遗传算法
权证定价
BP Artificial Neutral Networks
RBF Artificial Neutral Networks
Genetic Algorithm
Warrant Pricing
