文献类型：期刊文献

中文题名：在金融科技中基于人工智能算法的风险特征因子筛选框架的建立和在期货价格趋势预测相关的特征因子刻画的应用

英文题名：The Framework of Extract for Related Risk Factors by Using AI Algorithms and Applications to the Forecast of Trend for Commodity Futures Prices in Practice

作者：袁先智[1,2,3];周云鹏[3];刘海洋[3];严诚幸[3];钱国骐[4];钱晓松[5];汪冬华[6];李志勇[7];李祥林[8];林健武[9];沈思丞[3];曾途[3]

机构：[1]成都大学商学院,四川成都610106;[2]上海立信会计金融学院金融科技学院,上海201620;[3]成都数联铭品科技有限公司,四川成都610000;[4]墨尔本大学数学与统计学院,澳大利亚墨尔本VIC3010;[5]苏州大学金融工程研究中心,江苏苏州215000;[6]华东理工大学商学院,上海200237;[7]西南财经大学金融学院,四川成都611137;[8]上海高级金融学院,上海200030;[9]清华大学深圳国际研究生院,广东深圳518057

年份：2020

卷号：35

期号：4

起止页码：1

中文期刊名：安徽工程大学学报

外文期刊名：Journal of Anhui Polytechnic University

基金：国家自然科学基金资助项目(U1811462;71971031)。

语种：中文

中文关键词：大数据;吉布斯(Gibbs)随机搜索算法;特征筛选;关联方;价格趋势变化

外文关键词：big data;Gibbs sampling;stochastic search;feature extraction;related parties;forecast of price trend

摘要：研究的目的是建立对影响大宗商品期货价格变化趋势的关联风险特征因子的提取框架和配套的推断逻辑原理。具体来讲,以金融科技中大数据概念为出发点,利用人工智能中的吉布斯随机搜索(Gibbs Sampling)算法为工具,全面地陈述如何提取高度关联大宗商品期货价格变化的风险特征因子的流程和配套的逻辑原理,即采用(在马尔科夫链蒙特卡洛(MCMC)框架下)人工智能中的吉布斯随机抽样算法,结合OR值(Odds Ratio)作为关联分类和验证标准,实现从大量风险因子的数据中提取与大宗商品期货(铜)价格趋势变化相关的特征因子并进行分类,从而可用于构建支持期货价格趋势变化分析的特征指标。实证分析结果表明,该特征提取方法能够比较有效地刻画大宗商品期货(铜)价格的趋势变化,为业界进行大宗期货交易和风险对冲的管理提供了一种新的分析维度。另外,从影响价格趋势变化的特征因子中筛选出高度关联的特征指标的大数据分析方法,是与过去文献中对价格趋势分析的不同之处和创新点。
Based on the available macro and micro factors,this paper creates a massive pool of original data of both structured and unstructured factors and analyzes the effects of the factors from the pool on the change of the price of the relevant commodity futures.The method of Gibbs sampling of logistic regression candidate models is used for effective and scalable screening of all features,resulting in identifying those macro and micro-factors,with importance weighting measures,that influence the commodity price change.The empirical results show that the Gibbs sampling induced big data feature extraction algorithm can effectively extract the features related to the price trend of the Shanghai copper index contract.Further analysis of the associations between the trend of price changes and the identified effecting features reveals an important and sensible explanation of the social and business environment of the feature mapping of the futures association.The paper points out that our method of risk features extraction based in the big data framework is not only an innovation in theory for depicting the copper futures price trend,it also has technical innovations providing effective guidance in future copper trading in industrial practice.