Evolving traders and the business school with genetic programming: A new architecture of the agent-based artificial stock market (original) (raw)
Related papers
Genetic programming in the agent-based artificial stock market
Proceedings of the 1999 Congress on Evolutionary Computation-CEC99 (Cat. No. 99TH8406), 1999
In this paper, we propose a new architecture to study arti cial stock markets. This architecture rests on a mechanism called school" which i s a procedure to map the phenotype to the genotype or, in plain English, to uncover the secret of success. We propose an agent-based model of school", and consider school as an evolving population driven by single-population GP SGP. The architecture also takes into consideration traders' search behavior. By simulated annealing, traders' search density can be connected to psychological factors, such a s peer pressure or economic factors such a s t h e standard of living. This market architecture was then implemented in a standard arti cial stock market. Our econometric study of the resultant arti cial time series evidences that the return series is independently and identically distributed iid, and hence supports the e cient market hypothesis EMH. What is interesting though is that this iid series was generated by traders, who do not believe in the EMH at all. In fact, our study indicates that many of our traders were able to nd useful signals quite often from business school, even though these signals were short-lived.
Journal of Economic Behavior & Organization, 2002
By studying two well known hypotheses in economics, this paper illustrates how emergent properties can be shown in an agent-based artificial stock market. The two hypotheses considered are the efficient market hypothesis and the rational expectations hypothesis. We inquire whether the macrobehavior depicted by these two hypotheses is consistent with our understanding of the microbehavior. In this agent-based model, genetic programming is applied to evolving a population of traders learning over time. We first apply a series of econometric tests to show that the EMH and the REH can be satisfied with some portions of the artificial time series. Then, by analyzing traders' behavior, we show that these aggregate results cannot be interpreted as a simple scaling-up of individual behavior. A conjecture based on sunspot-like signals is proposed to explain why macrobehavior can be very different from microbehavior. We assert that the huge search space attributable to genetic programming can induce sunspot-like signals, and we use simulated evolved complexity of forecasting rules and Granger causality tests to examine this assertion.
Price Discovery in Agent-Based Computational Modeling of Artificial Stock Markets
2000
This paper studies the behavior of price discovery within a context of an agent based stock market, in which the twin assumptions ,n amely,rational expectations and the representative agents normally made in mainstream economics, are removed. In this model, traders stochastically update their forecasts by searching the business school whose evolution is driven by genetic programming .V ia these agent
Proc. of the 12th International Conference on Computing in Economics and Finance 2006, 2006
This paper investigates the process of deriving a single decision solely based on the decisions made by a population of experts. Four different amalgamation processes are studied and compared among one another, collectively referred to as central decision makers. The expert, also referred to as reference, population is trained using a simple genetic algorithm using crossover, elitism and immigration using historical equity market data to make trading decisions. Performance of the trained agent population’s elite, as determined by results from testing in an out-of-sample data set, is also compared to that of the centralized decision makers to determine which displays the better performance. Performance was measured as the area under their total assets graph over the out-ofsample testing period to avoid biasing results to the cut off date using the more traditional measure of profit. Results showed that none of the implemented methods of deriving a centralized decision in this investigation outperformed the evolved and optimized agent population. Further, no difference in performance was found between the four central decision makers.
Proc. of the IEEE/WIC/ACM International Conference on Intelligent Agent Technology, 2004
This paper investigates the effectiveness of an agent based trading system. The system developed employs a simple genetic algorithm to evolve an optimized trading approach for every agent, with their trading decisions based on a range of technical indicators generating trading signals. Their trading pattern follows a simple fitness function of maximizing net assets for every evolutionary cycle. Their performance is analyzed compared to market movements as represented by its Index, as well as investment funds run by human professionals to establish a relative measure of success. The results show that the developed system performs comparably to its human counterparts across different market environments, despite these agents being rather primitive in nature. Future forthcoming work will refine and explore the potential of this approach further.
Proc. of the 6th International Conference on Recent Advances in Soft Computing 2006, 2006
This paper investigates genetic algorithm evolved agents trading on real historical equity market data using technical analysis, the capital asset pricing model and a hybrid model of the two approaches. Three agent groups are generated, each using solely one of the two approaches or their hybrid to determine trading decisions. Each group consists of ten independently evolved populations over a thousand generations, whose elite’s performances are consequently averaged and used to compare to the other approaches. Results indicated that the technical analysis based approach performed better than the capital asset pricing model based approach, while the hybrid approach in turn outperformed both. As part of ongoing research, the results would suggest significant benefits in performance oriented implementation of hybrid or multi-method based approaches in agent-based systems.
Genetic Programming and Financial Trading: How Much About "What We Know
Springer Optimization and Its Applications, 2008
Based upon its performance in eight stock markets and eight foreign exchange markets during three consecutive test periods, this paper gives a thorough analysis of the application of genetic programming (GP) to financial trading. We reexamine a number of interesting findings obtained in some earlier studies, regarding both trading returns and trading behavior. Our findings are comparable to those earlier studies, and, to some extent, consolidate them. Nonetheless, by using a larger set of time series data, we are able to find evidence of how GP performance can be connected to the trending and cyclical properties of financial data. The specification of the data not only affects the profitability of the GP-discovered trading programs; it also determines the observed properties of the programs themselves. In this paper, we analyze the GP-discovered trading programs regarding their essence, complexity, activeness, and consistency. In light of these findings, we then highlight the progresses that could be made in the future. the two following concerns. 1 First, technical analysis generally does not refer to a fixed set of trading rules. They are evolving and changing over time. Many of them are still not even known to the public. However, for some time academic studies seem to have overlooked this property, and have tended to study them as if they are fixed over time. 2 It is, therefore, not surprising to see the diversity of the results: they are profitable in some markets some of the time, while they fail in other markets at other times, and so they are very inconclusive. 3 A more systematic way to study this evolving subject is to place it in a dynamic and evolving environment. Genetic programming, as a tool for simulating the evolution of trading rules in response to the changing environment, can then serve this purpose well. Second, technical analysis usually involves quite complicated transformations and combinations of price and volume signals, which is too demanding to be harnessed by the human mind. GP, as a rules-generating machine, can better facilitate us to travel trough this jungle.