A novel learning network for option pricing with confidence interval information

Abstract

Nonparametric approaches for option pricing have recently emerged as alternative approaches that complement traditional parametric approaches. In this paper, we propose a novel learning network for option-pricing, which is a nonparametric approach. The main advantages of the proposed method are providing a principled hyper-parameter selection method and the distribution of predicted target value. With these features, we do not need to adjust any parameters at hand for model learning and we can get confidence interval as well as strict predicted target value. Experiments are conducted for the KOSPI200 index daily call options and their results show that the proposed method works excellently to obtain prediction confidence interval and to improve the option-pricing accuracy.