Automatic Optimization Method On Option Pricing Model And Program Trading Based On GA/GP Technologies

In this paper, we use genetic algorithms (GA) and genetic programming (GP) optimization techniques, combined with financial theories, respectively, to achieve the parameters estimation of the option pricing models and automatic optimization of the program trading strategies.Genetic algorithms and genetic programming, are both global optimization algorithms, learning from the theory of evolution such as the rules of natural selection and survival of the fittest. Genetic algorithms apply to optimize the parameters of a particular model, with high efficiency, non-linear characteristics. Genetic programming, which is based on the genetic algorithm, achieved the op-timization of the model structure through the binary tree, thus greatly expanding the scope of its application in the field of machine learning and artificial intelli-gence.Combined with volatility estimation models and Option pricing models which are almost based on Black-Scholes formula, we can pricing the option products. Volatility estimation is the core of option pricing. In this paper, we study the estimation methods of Exponentially Weighted Moving Average model (EWMA), Generalized Autoregressive Conditional Heteroscedastic model (GARCH) and S-tochastic Volatility model (SV). We have designed a standard genetic algorithm function using the MATLAB Genetic Algorithm Toolbox, and developed a GA method for volatility model parameter estimation by programming different fit-ness functions. Moreover, with the combination of genetic algorithm and Monte Carlo simulation method, we have achieved the parameter estimation of GARCH-MC and SV models, and achieved the numerical option pricing, as an important part of the whole option prcing computer program. To verify the practicality of the genetic algorithm method, we use China warrants and Hong Kong’s Hang Seng Index Options data for empirical analysis. Considering from the results of the volatility estimation, GA methods for EWMA and GARCH models achieve the same accuracy with MATLAB built-in algorithm, and better portability than it. Empirical analysis shows the GARCH-MC method and SV model results in practical applications. Statistical arbitrage strategy based on Delta hedging is to verify the effect of option pricing. It shows that we can achieve a positive return in more accurate pricing varieties.In the second part of this paper, we discuss the use of genetic programming to achieve computer automatic optimization of the program trading strategies. Generally, it’s hard to develop a stable profit strategy for one person. We analyze the structure of technical analysis strategies, define the technical indicators as tree structures, and combine them into a complete strategy. Under the principle of genetic programming, We achieve the computer automatic optimization by optimize the performance targets of every strategy. Structural optimization of the strategy model is the mean characteristic of this method, which is different from the traditional parameter optimization.With the genetic programming principle, we completed the MATLAB-C mixed language program, including strategy construction, strategy back-testing and optimization modules. We use index futures data to test the effect of the program. After adjusting parameter settings for the data, we run the automatic optimization program, and analyze the effect of optimization. We give an example of a strategy codes which are developed by the computer, and discuss its trading logic and performance.