Real Options Valuation using Machine Learning Methods

Abstract

This paper follows work on establishing comprehensive framework for investment projects valuation discussed in [3], [8], [12], and [20]. Previous work is focused on capturing strategic value of investment projects while also incorporating strategic decisions of competitors. Two main methodologies that comprise such valuation framework are Real Options Analysis and Game Theory. In this paper, it is attempted to price real options using Machine Learning (ML) methods. First, selected machine learning models are trained to predict option prices as given by The Black-Scholes formula. Having shown some promise by work discussed in [22] and [23], real-world data has been selected for pricing options and then training machine learning models on them. Finally, various investment projects have been simulated to price option to expand using Cox-Ross-Rubinstein binomial model discussed in [4] and then train machine learning models to predict it. This, in turn, has potential to incorporate market competition implicitly in the value of the strategic option during training process. Hence, machine learning approach can become real options pricing method that is valid not only for monopolistic markets. With this aim, section 1 of the paper gives brief introduction of option pricing methods, section 2 uses Nasdaq Futures historical prices for training ML models to price financial options, and section 3 uses simulated investment projects for training ML models and pricin options to expand. Complete code is available at github.com/leongache/Real-Options-Valuation-using- Machine-Learning-Methods.