Options Pricing with Neural Networks
Session Number
Project ID: CMPS 09
Dr. Phadmakar Patankar, Illinois Mathematics and Science Academy
Discipline
Computer Science
Start Date
20-4-2022 11:25 AM
End Date
20-4-2022 11:50 AM
Abstract
In the modern day, Artificial Neural Networks (ANNs) have been recently used and tested to calculate and predict options pricing [Ke 2019], but currently, the most tested and reliable methods are the Binomial and Trinomial methods. Binomial and trinomial pricing models are valuation methods for financial derivatives, namely, options. The methods create a binomial or trinomial tree through an iterative process that allows for specification of nodes, or points of time, throughout the time span between the valuation date and an option’s expiration date [“How the Binomial,” 2020]. In both binomial and trinomial trees there is a starting option price, but the difference between the methods are the steps that the price can take. In a binomial tree, the price can either go up or down by a multiplier, but a trinomial tree also has a probability to stay constant [University of Leicester 2019].
However, these mathematical methods hold several flaws. The Binomial and Trinomial methods require an enormous amount of calculations when measuring a single optIn the modern day, Artificial Neural Networks (ANNs) have been recently used and tested to calculate and predict options pricing [Ke 2019], but currently, the most tested and reliable methods are the Binomial and Trinomial methods. Binomial and trinomial pricing models are valuation methods for financial derivatives, namely, options. The methods create a binomial or trinomial tree through an iterative process that allows for specification of nodes, or points of time, throughout the time span between the valuation date and an option’s expiration date [“How the Binomial,” 2020]. In both binomial and trinomial trees there is a starting option price, but the difference between the methods are the steps that the price can take. In a binomial tree, the price can either go up or down by a multiplier, but a trinomial tree also has a probability to stay constant [University of Leicester 2019].
However, these mathematical methods hold several flaws. The Binomial and Trinomial methods require an enormous amount of calculations when measuring a single option over a long period of time. Additionally, the Binomial method is far too simple. The underlying asset is forced to be worth two prices, which is inaccurate. ANNs have been shown to decrease times in calculations after their initial training period [Liu 2019].
In this study, we will be exploring how the calculations completed through trained ANNs compare against Binomial and Trinomial methods in predicting European and American Option prices. To do this, we will be using a dataset of American and European options from the time period of January to December 2019, specifically the put options of Alphabet, Inc. in the American Markets and Royal Dutch Shell in European Markets. And then they will be compared using real-time data for both. ion over a long period of time. Additionally, the Binomial method is far too simple. The underlying asset is forced to be worth two prices, which is inaccurate. ANNs have been shown to decrease times in calculations after their initial training period [Liu 2019]. In this study, we will be exploring how the calculations completed through trained ANNs compare against Binomial and Trinomial methods in predicting European and American Option prices. To do this, we will be using a dataset of American and European options from the time period of January to December 2019, specifically the put options of Alphabet, Inc. in the American Markets and Royal Dutch Shell in European Markets. And then they will be compared using real-time data for both.
Options Pricing with Neural Networks
In the modern day, Artificial Neural Networks (ANNs) have been recently used and tested to calculate and predict options pricing [Ke 2019], but currently, the most tested and reliable methods are the Binomial and Trinomial methods. Binomial and trinomial pricing models are valuation methods for financial derivatives, namely, options. The methods create a binomial or trinomial tree through an iterative process that allows for specification of nodes, or points of time, throughout the time span between the valuation date and an option’s expiration date [“How the Binomial,” 2020]. In both binomial and trinomial trees there is a starting option price, but the difference between the methods are the steps that the price can take. In a binomial tree, the price can either go up or down by a multiplier, but a trinomial tree also has a probability to stay constant [University of Leicester 2019].
However, these mathematical methods hold several flaws. The Binomial and Trinomial methods require an enormous amount of calculations when measuring a single optIn the modern day, Artificial Neural Networks (ANNs) have been recently used and tested to calculate and predict options pricing [Ke 2019], but currently, the most tested and reliable methods are the Binomial and Trinomial methods. Binomial and trinomial pricing models are valuation methods for financial derivatives, namely, options. The methods create a binomial or trinomial tree through an iterative process that allows for specification of nodes, or points of time, throughout the time span between the valuation date and an option’s expiration date [“How the Binomial,” 2020]. In both binomial and trinomial trees there is a starting option price, but the difference between the methods are the steps that the price can take. In a binomial tree, the price can either go up or down by a multiplier, but a trinomial tree also has a probability to stay constant [University of Leicester 2019].
However, these mathematical methods hold several flaws. The Binomial and Trinomial methods require an enormous amount of calculations when measuring a single option over a long period of time. Additionally, the Binomial method is far too simple. The underlying asset is forced to be worth two prices, which is inaccurate. ANNs have been shown to decrease times in calculations after their initial training period [Liu 2019].
In this study, we will be exploring how the calculations completed through trained ANNs compare against Binomial and Trinomial methods in predicting European and American Option prices. To do this, we will be using a dataset of American and European options from the time period of January to December 2019, specifically the put options of Alphabet, Inc. in the American Markets and Royal Dutch Shell in European Markets. And then they will be compared using real-time data for both. ion over a long period of time. Additionally, the Binomial method is far too simple. The underlying asset is forced to be worth two prices, which is inaccurate. ANNs have been shown to decrease times in calculations after their initial training period [Liu 2019]. In this study, we will be exploring how the calculations completed through trained ANNs compare against Binomial and Trinomial methods in predicting European and American Option prices. To do this, we will be using a dataset of American and European options from the time period of January to December 2019, specifically the put options of Alphabet, Inc. in the American Markets and Royal Dutch Shell in European Markets. And then they will be compared using real-time data for both.