Building a Neural Network to Predict the Option Price

With the increase in the financial literacy of the population, the scale of the financial market is also expanding: in 2021, the number of private investors who opened brokerage accounts on the Moscow Exchange almost doubled compared to 2020 and at the beginning of 2022 is more than 17 million. One of the most effective tools for reducing market risks are various derivative financial instruments. The aim of the study is to improve the quality and efficiency of estimating the value of an option on the index of the Russian trading system by developing and implementing a specialized information system. To achieve this goal, the following tasks were set and solved in the work: 1) an analysis of the main concepts, tools and algorithms for evaluating options using machine learning methods was carried out; 2) the components of a deep learning model for valuing an option on the RTS Index are determined; 3) a statistical interpretation of the processed data was carried out, 4) a neural network was built for put and call options. Materials and methods. Statistical analysis and neural networks apparatus were used in modeling. Conclusions. A study was made of the statistical characteristics of the underlying asset on the RTS Index futures; an algorithm was developed that uses the fair value of money indicator RUSFAR, calculated by the Moscow Exchange, instead of using zero-coupon rates, which are biased due to the incomplete backing of funds by assets to assess risk-free borrowing rates; the obtained results of the models are interpreted and conclusions are formulated regarding the quality of the obtained models.

Financial market, options execution, neural networks, loss function, option value models