圣塔菲人工股票市场java代码

# Source All code and material is drawn from http://artstkmkt.sourceforge.net This repo is a conversion of the CVS repository available at that address. # Purpose SFI-ASM is a captivating, enigmatic thing. My interest is first the preservation of the historical artifact, second the understanding of them model, and third the understanding of the academic-historical context of the model's creation. To that end, this repository tracks the source code through all of the releases I can find, and will try to modify the code and codify its environment sufficiently s.t. it can be built on a well specified VM. # SFI-ASM The Sante Fe Institute Artificial Stock Market is a description of a agent-based model of a stock market. It has been realized in a few different implementations, which this repository aims to collect and enhance. The model itself was first described by Palmer et al. (1994) ## Artstkmkr ``` Group/Project Resource Page lists all resources available for this project. Several articles* have been published using the Artificial Stock Market model. The model simulates prices & trade levels in a market made up of "artificial adaptive agents." This is a prototype example of a "complex system" and is thought to illustrate the benefits of simulation modeling. This page is dedicated to the revision/updating/enhancement of the Swarm version of the ASM. The first objective was to bring the ASM code--as released in April, 2000, with the version label 2.0--up to date and into conformance with Swarm programming style and recommendations for Swarm 2.1. That was achieved with ASM-2.2, which represents a substantially reworked and enhanced version of the model. Actually, ASM-2.2 went a bit further than that, because it (optionally) uses some hdf5 data storage features that were supported only in Swarm after 2.1.1 (prerelease snapshots of Swarm-2.2). The next step, ASM-2.4, includes the addition of a number of new features, including serialization (ability to stop and restart a model). ASM-2.4 now performs, as far as Thomas Badegruber and I can tell, in a way that replicates the original Santa Fe Stock Market. See the Swarm community links page for release info. In case you wonder what a Swarm program looks like while it runs, I have a directory of screenshots!, such as screenshots/asm-20000530.gif . The newest version, ASM-2.4, looks like this: screenshots/asm-20030606.png on my RedHat Linux system. ``` # References Arthur, W. Brian, John H. Holland, Blake LeBaron, Richard G. Palmer, and Paul Tayler. "Asset pricing under endogenous expectations in an artificial stock market." Available at SSRN 2252 (1996). Ehrentreich, Norman. "A corrected version of the santa fe institute artificial stock market model." Complexity (2003). Ehrentreich, Norman. Agent-based modeling: The Santa Fe Institute artificial stock market model revisited. Vol. 602. Springer Science & Business Media, 2007. LeBaron, Blake, W. Brian Arthur, and Richard Palmer. "Time series properties of an artificial stock market." Journal of Economic Dynamics and control 23, no. 9 (1999): 1487-1516. LeBaron, Blake. "Building the Santa Fe artificial stock market." Physica A (2002). Palmer, Richard G., W. Brian Arthur, John H. Holland, Blake LeBaron, and Paul Tayler. "Artificial economic life: a simple model of a stockmarket." Physica D: Nonlinear Phenomena 75, no. 1 (1994): 264-274. # Original Readme (ASM-2.4): ``` Swarm ASM notes. If you are looking for highlights, this is the right spot. Paul Johnson <pauljohn@ku.edu> ASM-2.4 June 2003 This version requires Swarm-2.1.143 or newer. If you don't have that, you will have to disable serialization features and compile without using HDF5 data output. Thomas Badegruber <thomas.badegruber@uni-graz.at> visited here in Kansas in March and then went to Boston to work with Blake LeBaron. That re-energized me on this project. Thomas is now a full contributor to the CVS archive on Sourceforge. We have made changes to improve functionality, fix bugs, and make the genetic algorithm code more easy to read and extend. And, perhaps most importantly, we can now re-produce the finding that there is a major difference in the usage of technical and fundamental bits when the agent use of the genetic algorithm is made more frequent. That is to say, when agents use the GA every 250 periods, they are substantially more likely to use technical trading information than if they update every 1000 periods. Because this behavior is now reproduced, we are confident that the Swarm-ASM behaves in the same way in all substantial ways as the Next/Objective-C models presented in the literature. Functionality enhancements * Graph showing usage of funamental bits, technical bits, and dummy bits full model serialization. * Serialization: Simulation can be saved in a "snapshot" and restarted. * Parameter values have been reset to match the ones used in publications. * Uses enhancements in Swarm's EZGraph (changes introduced since Swarm-2.1.1) to save output data as vectors in HDF5 files. This significantly enhances usage of output data files in statistical analysis programs. Bug fixes / code changes (see ChangeLog for details, or persue the patch from ASM-2.2 to ASM-2.4.) * closed memory leakes caused by use of probes in the getDouble() function in BFParams.m. * revised some of the scheduling apparatus to make the model run faster. * Fixed a misplaced bracked in BFagent's MakePool method. This was discovered and corrected by Thomas Badegruber. It significantly affects the performance of the genetic algorithm over the long run. * Coding for graphs and data output of hdf5 vectors for time plots was reorganized, significanly simplifying the shift from GUI to nonGUI models. Look in Output.m ASM-2.2 November 2001 This work was mainly aimed at documentation, inserting Autodoc markers so I can build nicer looking documentation. The only big substantive cleanup was in ASMModelSwarm. I had been avoiding this for a long time because the scheduling that was used was very complicated and hard to understand. I succeeded in a major simplicification and cleanup and have verified the numerical results are identical. This is explained in the beginning of ASMModelSwarm.m There was also a "procedural" cleanup of the code in World.m. The old version had a lot of callocs and other memory magic because it was keeping moving averages "by hand". This has been changed. There is now a MovingAverage class that is used to do that magic, so the code in World.m is much easier to read. There is much less low level math floating around. SNAP3: ASM 20011024 October 2001 The aim has been to bring this model into convenient Swarm standards. Actually, the real aim has been to make ASM an effective teaching tool and an example of good simulation coding practice. Part of that is a modernization of the data handling, both the input of parameter values, and record keeping of parameter values, as well as the saving of data. One of my pet peeves about Swarm is that we do not have a standard, idiot proof way to keep records on the settings and results are for a particular run. We also don't have fool proof ways to interate models and run simulations over again. The first problem is dealt with here in the Output class, and the second one is dealt with by the creation of a Parameter class, where we could manage command line parameters if we wanted to. 1. Inputting Parameter Values. First, note the big parameter input file "asm.scm". This has the initial values for three classes, bfParams, asmModelParms, and asmBatchParams. This data is "called" to create the classes, you have to stare at it a while to get used to it. Where we used to have the createBegin/createEnd routine for a class, now we've got this different thing that creates a class and initializes its ivars for us, like so asmModelParams =[lispAppArchiver getWithZone: self key: "asmModelParams"]) The asmModelParms cla