| Because of the lack of the controllable experimental condition under which the conclusions could be reproduced, traditional economic and financial theories are hard to be recognized as science. In order to satisfy the applicable conditions of Probability and Statistics, the Probability and Statistics based Modern Capital Market Theory (MCMT), was built on three hypotheses: rational investor, efficient market, and the random walk of yield rate. However, there are many doubts about these hypotheses, which come from the investors'psychology and behaviors analyses, rules and realities of the real capital market, and the positive analyses about the real financial time series. Above all, the lack of financial experimental conditions causes the quantization analyses in financial field still belong to empiricism methodology.After great development for decades, the Scientific Computation becomes the third kind of scientific research methodology, succeeding the Scientific Theory and the Scientific Experiment. Many problems, which are traditionally difficult to be proved by theoretical derivation or experimental verification, could be explored by computational simulation approaches. Since 1950s, the complexity science and the complex system modeling methods have been developed, and have become the powerful tools to transform the real problems to computational models. The application of the complexity theories and the complex system modeling methods to economic and financial fields brings a new ray of hope for the research on the internal dynamical mechanisms of financial markets. Especially, the Artificial Financial Market models, led by the Santa Fe Institute Artificial Stock Market (SFI-ASM), opened up a brand-new kind of research methodology in Experimental Finance Field. With the help of the powerful describing abilities of the complex system modeling methods, the hypotheses, concepts, and conclusions of new related fields, such as the Behavioral Finance and the Psychology of Finance, could be expressed or verified in the artificial financial market models. Then, these new theories and the experimental finance could be promoted mutually.The mainstream artificial financial markets like SFI-ASM, are based on the Multi-Agent modeling technology. The Multi-Agent models are advanced in describing heterogeneous individual investors'behaviors. But they lack the ability to express the interactions within the investors. Herd behavior, information spreading, and individuals'interactions, however, are ubiquitous in real capital market. The Cellular Automata, benefiting from their formalized normal description method about correlations within individuals, can cover the shortage of the Multi-Agent technology. So the cellular automata become the foundation of another kind of artificial financial markets. However, because most pre-existing cellular automata based artificial financial markets. Most pre-existing cellular automata based artificial financial markets, however, adopted classiccellular automata's topology as their neighborhood structure. This caused the relations within individuals in these models are given apriori. Furthermore, these models described the details in capital market structure imperfectly. So, their maturity could not be compared with multi-agent based artificial financial markets like SFI-ASM.The artificial financial markets have great prospects in exploring financial complex system, but the pre-existing cellular automata based artificial financial markets still have shortages mentioned above. On the basis of cellular automata theory, we studied the financial complex system's theory, modeling methods, experimental technology, analysis methods of experimental data etc. By means of abstracting the operating mechanism of real capital markets, we built a complete artificial financial market modeling framework, which was named as Emergence-Artificial Financial Market Framework, E-AFMF. E-AFMF could provide tools for financial complex system theories'tests and verifications.First of all, we extended the classic definition cellular automata to an abstract one with network topology and heterogeneous individual cells. And we promoted the formal definition of it. Next, on the basis of C++ template and Microsoft's Concurrency Runtime technology, we realized a universal Cellular Automata Parallel Template Library, CAPTL. On the models'description ability side, a series of technologies are adopted in CAPTL to provide max flexibility for various cellular automata. These include C++ template technology to parameterize cell states'type; abstract base class of cell objects, which is used to build an inheritance hierarchy to realize heterogeneous cell individuals in a concrete model; and network topology which is used to represent neighborhoods of cellular automata, so that any kind of discrete structure could be expressed in a model. On the models'simulation technology side, a Concurrency Runtime based thread level parallel model of cellular automata has been realized in CAPTL. Any CAPTL based cellular automata model could be simulated in shared memory parallel machines.E-AFMF is built on the basis of CAPTL. Investors in artificial financial market are realized as a heterogeneous cellular automata model with a network neighborhood. Meanwhile, we abstracted the organization structure, operation mechanism, trading process, pricing mechanism of the real capital market, the information transmission model and the interaction pattern within investors reasonably. And we realized a corresponding framework for these concepts in E-AFMF. Extended from the fine-grained parallel model in CAPTL, a coarse-grained parallel model was added into E-AFMF to realize other modules except investors in the artificial financial market. Then, the E-AFMF formed an adaptive feedback-iterated modeling framework for the complex dynamic systems. Benefiting by the E-AFMF, we can customize a brand new artificial financial market model flexibly, only if we defined the behavior rule of investors, the information's form and transmission model. The purpose of building an artificial financial market model is to verify the relationship between the macroscopic dynamic feature and the microcosmic inner structure of the capital market. Any E-AFMF based artificial financial market model could produce a price-volume time series, which has the same form with the price-volume time series of the real capital market, and several indicator time series, which are about the inner structure of the investors in the artificial financial market model. The price-volume time series produced by artificial financial market models could be analyzed in the same way as which produced by the real capital markets. Meanwhile, we can we can comparatively study the price-volume time series and the indicator time series of the investors inner structure. So, the E-AFMF based artificial financial market models could become powerful research tools combining financial theory experiment and empirical analyses.The E-AFMF is a creative trial, in which the complex system modeling theories, methods, and parallel simulation technology are applied to the experimental finance field. The in-depth development of E-AFMF could bring a brand new research method for the capital market theory. Furthermore, the research achievements of the E-AFMF based artificial financial market models could be valuable for the risk warning and management of the real capital market.
|
PDF Full Text Request