摘要
预测市场是经济学里用以分析和预见经济发展趋势的技术,这篇论文研究如何将预测市场作为机器学习的一种工具。论文概述了当前预测市场领域的研究和进展、并将预测市场和当前的模型组合方法关联起来,展示如何利用它们。然后基于现有理论,论文研究并实现了一个人工预测市场,同时利用实际数据对它的表现进行测试和评估,得出预测市场应用于机器学习领域的作用与效果,根据评估结果来理解预测市场技术用于机器学习的优势和弱点,从而得出进一步的研究方向。
The main goal of the thesis is to explore the possibility of using prediction markets to improve machine learning. Firstly, the thesis gives a brief introduction of present situation of this area, relates artificial predictive markets to currently famous model combination techniques and shows how to extend them. It also develops techniques which was in-troduced first in a previously known framework, includes a description of how to implement this framework and how to evaluate its performance on both synthetic and real data sets. Finally, with the help of the results of this evaluation, it ana-lyses the strengths and weaknesses of this approach and makes many useful conclusions and suggests future directions in this research area.
出处
《软件》
2014年第11期31-35,共5页
Software
关键词
机器学习
预测市场
分类算法
模型组合
machine learning
artificial prediction markets
classification algorithm
prediction markets
model combination