A new hybrid parametric and machine learning model with homogeneity hint for European-style index option pricing精品.pdfVIP

Neural Comput Applic DOI 10.1007/s00521-016-2303-y ORIGINAL ARTICLE A new hybrid parametric and machine learning model with homogeneity hint for European-style index option pricing 1 2 Shom Prasad Das • Sudarsan Padhy Received: 8 September 2015 / Accepted: 30 March 2016 The Natural Computing Applications Forum 2016 Abstract Here, we propose and investigate a hybrid compared with the benchmarked BS model. We observe model that combines parametric option pricing models similar improvements over the other benchmarked models. such as Black–Scholes (BS) option pricing model, Monte Therefore, the proposed new hybrid model is a suit- Carlo option pricing model, and finite difference method able alternative model for option pricing when higher with nonparametric machine learning techniques such as predictive accuracy is desired. support vector regression (SVR) and extreme learning machine-based regression models. The purpose of this Keywords Option pricing Support vector regression model is to support better investment decisions by fore- (SVR) Extreme learning machines (ELMs) casting the option price with high predictive accuracy. To Homogeneity hint Nonparametric methods Parametric further reduce the forecasting error, we incorporate a methods homogeneity hint (i.e., training the model by categorizing the options data based on moneyness and time-to-maturity of the option contract) into the model. We examine the 1 Introduction feasibility and effectiveness of this model using a case study to predict the one-day-ahead price of index options Options are significant financial derivative instruments that traded in the National Stock Exchange of India Limited. allow investor