PRICING EUROPEAN AND AMERICAN OPTION AS BI-OBJECTIVE OPTIMIZATION PROBLEM WITH NEURAL NETWORK ALGORITHM (NNA)

• Main Author: Masli, Aziza
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/80791
• Institution: Institut Teknologi Bandung

An option is a type of financial derivative that can help investors avoid the risks that may arise during investing. An investor undoubtedly seeks the accurate value of an option before purchasing a specific option contract. Numerous techniques have been introduced to solve the problem of pricing the option value; however, there are still shortcomings and limitations in capturing the real market conditions. In this study, the option pricing problem is modeled as a bi-objective optimization problem, with payoff and the probability of obtaining that payoff as the respective objective functions. Furthermore, the option value, initially modelled as a biobjective optimization problem, will be transformed into a single objective optimization problem using the Adaptive Weighted Sum Method. To solve the option’s pricing problem, we will use Neural Network Algorithms that was newly introduced in 2018. Using historical data, we can determine the option value of an underlying asset in the subsequent period. The result show that the Neural Network Algorithms can capture the optimal pareto solutions to approach the prices of European and American options. It is shown by the less errors obtained with the Black-Scholes Differential Equation, NSGA-II and Binomial Methods.