YucongDuan的个人博客分享 http://blog.sciencenet.cn/u/YucongDuan

博文

AGI-AIGC-GPT测评DIKWP(全球)实验室(测评系列之十一)-世界上第一篇金融领域研究报告

已有 2553 次阅读 2023-4-19 22:33 |系统分类:论文交流

The Capability Evaluation of GPT-4 and Tongyi Qianwen on Financial Domain with DIKWP Analysis (AGI-AIGC-GPT Test and Evaluation DIKWP (Global) Lab)

April 2023

DOI: 10.13140/RG.2.2.12604.92804

https://www.researchgate.net/publication/370112349_The_Capability_Evaluation_of_GPT-4_and_Tongyi_Qianwen_on_Financial_Domain_with_DIKWP_Analysis_AGI-AIGC-GPT_Test_and_Evaluation_DIKWP_Global_Lab

AGI-AIGC-GPT测评DIKWP(全球)实验室

(测评报告系列之十一)

GPT-4 通义千问的金融领域DIKWP能力评测报告

Fuliang Tang1, Yucong Duan1,2*

22210812000006@hainan.edu.cn1, duanyucong@hotmail.com1,2 

AGI-AIGC-GPT Test and Evaluation DIKWP (Global) Lab1

DIKWP research group, Hainan University2

1. 摘要

随着大型语言模型LLMs)的大量涌现,当前通用人工智能(AGI)和GPT技术的发展引起了广泛关注。目前在 AGI/GPT 问题上,针对大型语言模型生成能力的的评价测试方法大多基于若干封闭任务、评价者 主观感受或者认知经验,缺乏全面覆盖、评价客观和过程统一的针对AGI/GPT场景的评价框架和评测标准。尤其是大语言模型在金融领域的测试并不是特别的全面。本文旨在探讨如何客观有效地评价通用人工智能(AGI)在金融领域的智能水平和能力范围。本文基于数据、信息、知识、智慧和智能(DIKWP)模型,构建了一个统一的AGI/GPT金融测试体系,并选取了目前最先进的AGI模型GPT-4和通义千问作为评价对象。本文以量化交易为切入点,测试GPT-4以及通义千问在策略生成方面的能力,以及在DIKWP模型下GPT-4和通义千问对量化交易类问题的处理能力。本文为AGI/GPT金融测试提供了一个新的视角和方法,并为未来AGI/GPT模型的改进和发展提供了有益参考。

 

Keywords: GPT-4, Tongyi Qianwen,DIKWP, AGI, Financial, Quantitative Trading, LLM

2. 简介

人工智能(AI)领域的一个终极目标是通用人工智能(AGI)[1],也就是能够在任何人类专业领域内展现出与人类相当或超越的智能水平的AI。近年来,AGI相关的技术在文本处理、机器翻译、图片处理等方面取得了巨大的进步,极大地提升了人类对数字资源的意图理解能力。然而,如何客观有效地评估AGI在金融领域的智能水平和能力范围仍然是一个难题。现有的AGI评价测试方法大多依赖于主观认知经验,缺乏一个统一的评价体系和评测标准。本研究旨在建立一个基于数据、信息、知识、智慧和意图(DIKWP)模型的AGI金融测试体系本文的特色是以量化交易为切入点,分别让GPT-4和通义千问独立处理量化交易问题、在DIKWP模型支撑下处理量化交易问题以及让其处理歧义量化交易问题。本研究为AGI评价测试提供了一个新的视角和方法,并为未来AGI模型的改进和发展提供了有益参考。

3. GPT-4的金融相DIKWP用示例

我们将首先GPT-4在没有DIKWP支撑的前提下让其独立处理量化交易问题。然后GPT-4运用DIKWP的框架处理量化交易问题,进一步的我们给出不同人的对具体实例的不同量化交易策略(歧义),让GPT-4基于DIKWP给出最合理的量化交易策略。最后,我们将三种所获得的回答结果进行比较。

3.1 量化交易的GPT-4处理

 

Prompt:

 

1、中微公司(183.940, -6.94, -3.64%)公告,巽鑫投资拟减持不超2%公司股份。

2、据港交所文件,新加坡政府旗下的投资机构GIC Private Limited412日出售了2038万股中国电信(7.080, 0.05, 0.71%)港股,平均减持价格为每股4.3811港元,持股比例下降至9.99%

3、药明康德(80.820, -4.80, -5.61%)(维权)公告,实控人控制的股东等拟减持不超3%股份。

4、海天瑞声(189.600, 1.85, 0.99%)公告,唐涤飞拟减持不超3.57%公司股份。

5、歌尔股份(20.300, -2.26, -10.02%)公告,2023年一季度净利润1.06亿元,同比下降88.22%

6、德业股份(281.850, 15.80, 5.94%)披露年报,2022年净利同比增162.28%,拟10822.6元。

7、金山办公(441.550, 14.28, 3.34%)公告,控股股东及奇文N维拟减持不超过4.735%

8、亿纬锂能(67.860, -1.11, -1.61%)公告,2022年净利润同比增长21%,拟101.6元。

9、首都在线(18.950, 0.90, 4.99%)在互动平台表示,正与燧原科技开展基于大模型的MaaS服务的联合研发。

根据上述内容,决定应买进哪家公司的股票,并给出详细过程

GPT-4回答:

 

根据上述信息,以下是对各公司的简要分析:

1. 中微公司:巽鑫投资计划减持不超过2%的公司股份,这可能导致短期内股价承压。

2. 中国电信:新加坡政府旗下投资机构GIC已经减持了一部分股份,这可能对股价产生一定影响。

3. 药明康德:实控人控制的股东拟减持不超过3%的股份,可能导致短期股价承压。

4. 海天瑞声:唐涤飞计划减持不超过3.57%的公司股份,可能导致短期股价承压。

5. 歌尔股份:2023年一季度净利润同比下降88.22%,表现较差。

6. 德业股份:2022年净利润同比增长162.28%,拟10转8派22.6元,公司业绩较好且分红政策较为慷慨。

7. 金山办公:控股股东及奇文N维拟减持不超过4.735%,可能导致短期股价承压。

8. 亿纬锂能:2022年净利润同比增长21%,拟10派1.6元,公司业绩稳定,分红政策也较为合理。

9. 首都在线:正与燧原科技开展基于大模型的MaaS服务的联合研发,展示了公司在技术领域的合作和发展潜力。

综合以上信息,我建议购买德业股份(6)的股票。原因如下:

1. 2022年净利润同比增长162.28%,显示出公司良好的盈利能力和发展势头。

2. 10转8派22.6元的分红政策较为慷慨,对于投资者来说具有较高的吸引力。

3. 相较于其他公司的减持、业绩下滑等负面因素,德业股份的整体表现更为稳定且具有成长潜力。

请注意,以上建议仅基于所提供的信息进行分析,实际投资风险请自行评估。建议在进行投资决策前,充分了解公司基本面和市场情况,并咨询专业投资顾问。

 

Prompt:

 

请结合上面的内容给合适的出量化交易策略

GPT-4回答:

 

结合上述信息,我们可以为德业股份(6)制定一个简单的量化交易策略。这里我们假设投资者希望通过多因子模型进行投资决策。以下是一个基于动量、盈利和估值因子的简单量化交易策略:

1. 数据准备:收集德业股份过去一年的日线数据,包括收盘价、成交量、市盈率、市净率等。

2. 动量因子:计算股票的20日、60日、120日收益率,用于衡量短期、中期和长期的动量表现。我们将以60日收益率为例,计算方法为:(当日收盘价 - 60天前收盘价) / 60天前收盘价。

3. 盈利因子:基于公司的净利润增长率(在本例中为162.28%),评估公司的盈利能力。

4. 估值因子:使用市盈率和市净率评估公司的估值水平。较低的市盈率和市净率可能意味着公司被低估。

5. 标准化处理:对以上三个因子进行标准化处理,将它们转化为z-score,以消除量纲影响。

6. 综合评分:为每个因子分配权重,如动量因子30%,盈利因子40%,估值因子30%。计算加权后的综合评分。

7. 交易信号:当综合评分超过某一阈值时(如1.0),认为股票具有投资价值,可以买入;当综合评分低于另一阈值时(如-1.0),则认为股票没有投资价值,可以卖出或避免购买。

8. 仓位管理:根据交易信号和个人风险承受能力,决定买入或卖出的仓位。例如,可以将可用资金的50%用于购买德业股份,其余资金用于其他投资或现金储备。

9. 风险管理:设定止损点,如当股价跌破10日均线时止损。定期调整投资组合,确保风险分散。

10. 回测与优化:对策略进行历史数据回测,评估策略的有效性和稳定性。根据回测结果优化因子权重、信号阈值等参数,提高策略的盈利能力和抗风险能力。

请注意,以上策略仅为示例,实际应用时需要根据投资者的具

 

 

3.2 基于DIKWP的量化交易的GPT-4处理

 

Prompt:

 

1、中微公司(183.940, -6.94, -3.64%)公告,巽鑫投资拟减持不超2%公司股份。

2、据港交所文件,新加坡政府旗下的投资机构GIC Private Limited412日出售了2038万股中国电信(7.080, 0.05, 0.71%)港股,平均减持价格为每股4.3811港元,持股比例下降至9.99%

3、药明康德(80.820, -4.80, -5.61%)(维权)公告,实控人控制的股东等拟减持不超3%股份。

4、海天瑞声(189.600, 1.85, 0.99%)公告,唐涤飞拟减持不超3.57%公司股份。

5、歌尔股份(20.300, -2.26, -10.02%)公告,2023年一季度净利润1.06亿元,同比下降88.22%

6、德业股份(281.850, 15.80, 5.94%)披露年报,2022年净利同比增162.28%,拟10822.6元。

7、金山办公(441.550, 14.28, 3.34%)公告,控股股东及奇文N维拟减持不超过4.735%

8、亿纬锂能(67.860, -1.11, -1.61%)公告,2022年净利润同比增长21%,拟101.6元。

9、首都在线(18.950, 0.90, 4.99%)在互动平台表示,正与燧原科技开展基于大模型的MaaS服务的联合研发。

请抽取出其中的数据、信息、知识、智慧、意图,并以表格的形式输出

GPT-4回答:

 

 

首先给出基本信息提炼表格,如下:

序号公司名称股票代码当前价格涨跌幅度公告/财报内容1中微公司183.940-6.94%-3.64%巽鑫投资拟减持不超2%公司股份2中国电信7.0800.05%0.71%新加坡政府旗下投资机构减持股份3药明康德80.820-4.80%-5.61%实控人控制的股东等拟减持不超3%股份4海天瑞声189.6001.85%0.99%唐涤飞拟减持不超3.57%公司股份5歌尔股份20.300-2.26%-10.02%2023年一季度净利润同比下降88.22%6德业股份281.85015.80%5.94%2022年净利同比增长162.28%,拟10转8派22.6元7金山办公441.55014.28%3.34%控股股东及奇文N维拟减持不超过4.735%8亿纬锂能67.860-1.11%-1.61%2022年净利润同比增长21%,拟10派1.6元9首都在线18.9500.90%4.99%正在与燧原科技开展基于大模型的MaaS服务的联合研发

数据:股票代码、当前价格、涨跌幅度、减持股份的比例、净利润等

信息:公告时间、股东信息、业务合作

知识:公司财务状况、股东减持规模与比例

智慧:结合公司财务状况和股东减持规模,判断股票是否值得投资

意图:提供投资决策参考

进一步提炼出数据、信息、知识、智慧、意图的具体内容如下:

数据:

· 中微公司减持股份的比例为不超过2%;

· GIC Private Limited 减持了中国电信股份,持股比例下降至9.99%;

· 药明康德实控人控制的股东等拟减持不超3%股份;

· 海天瑞声唐涤飞拟减持不超3.57%公司股份;

· 歌尔股份2023年一季度净利润为1.06亿元,同比下降88.22%;

· 德业股份2022年净利同比增长162.28%,拟10转8派22.6元;

· 金山办公控股股东及奇文N维拟减持不超过4.735%;

· 亿纬锂能2022年净利润同比增长21%,拟10派1.6元;

· 首都在线正与燧原科技开展基于大模型的MaaS服务的联合研发。

信息:

· 公司减持股份;

· 公司发布的财务报告,包括净利润等数据;

· 公司派发的股息;

· 公司的股东信息。

知识:

· 公司的财务表现,包括净利润同比增长或下降等;

· 公司股份变动情况,包括减持、增持等;

· 公司的股东情况,包括控股股东、实控人等。

智慧:

· 对公司股票的买卖决策,需要对公司的财务表现、股东情况等进行综合分析。

意图:

· 公司发布公告和财务报告,向投资者传达公司的股份变动情况、财务状况等信息;

· 投资者需要根据这些信息,做出买卖股票的决策。

将其以表格形式输出:

序号公司名称数据信息知识智慧意图1中微公司183.940, -6.94, -3.64%巽鑫投资拟减持不超2%公司股份实际控制人将减持减持可能对短期股价产生压力股东意图减持2中国电信7.080, 0.05, 0.71%GIC出售2038万股港股,持股比例下降至9.99%新加坡政府旗下投资机构已减持减持可能对股价产生一定影响政府投资机构意图减持3药明康德80.820, -4.80, -5.61%实控人控制的股东等拟减持不超3%股份实际控制人将减持减持可能对短期股价产生压力股东意图减持4海天瑞声189.600, 1.85, 0.99%唐涤飞拟减持不超3.57%公司股份实际控制人将减持减持可能对短期股价产生压力股东意图减持5歌尔股份20.300, -2.26, -10.02%2023年一季度净利润1.06亿元,同比下降88.22%公司业绩下滑业绩下滑可能导致投资者信心减弱报告公司财务状况6德业股份281.850, 15.80, 5.94%2022年净利同比增162.28%,拟10转8派22.6元公司业绩增长且分红政策慷慨良好的业绩和分红政策可能吸引投资者报告公司财务状况7金山办公441.550, 14.28, 3.34%控股股东及奇文N维拟减持不超过4.735%实际控制人将减持减持可能对短期股价产生压力股东意图减持8亿纬锂能67.860, -1.11, -1.61%2022年净利润同比增长21%,拟10派1.6元公司业绩增长且有分红计划良好的业绩和分红政策可能吸引投资者报告公司财务状况9首都在线18.950, 0.90, 4.99%与燧原科技开展基于大模型的MaaS服务的联合研发公司正在积极探索新技术领域技术合作可能提高公司竞争力和未来增长潜力报告技术合作及研发计划

 

Prompt:

 

结合上面输出的表格,分析这9支股票的优劣,并作出比较

GPT-4回答:

 

结合表格中的数据、信息、知识、智慧和意图,以下是对这9支股票的优劣分析:

1. 中微公司

优点:暂无明显优点

缺点:实际控制人减持,可能对短期股价产生压力

2. 中国电信

优点:暂无明显优点

缺点:新加坡政府旗下投资机构减持,可能对股价产生一定影响

3. 药明康德

优点:暂无明显优点

缺点:实际控制人减持,可能对短期股价产生压力

4. 海天瑞声

优点:暂无明显优点

缺点:实际控制人减持,可能对短期股价产生压力

5. 歌尔股份

优点:暂无明显优点

缺点:2023年一季度净利润同比下降88.22%,公司业绩下滑,可能导致投资者信心减弱

6. 德业股份

优点:2022年净利同比增162.28%,分红政策慷慨,良好的业绩和分红政策可能吸引投资者

缺点:暂无明显缺点

7. 金山办公

优点:暂无明显优点

缺点:实际控制人减持,可能对短期股价产生压力

8. 亿纬锂能

优点:2022年净利润同比增长21%,有分红计划,良好的业绩和分红政策可能吸引投资者

缺点:暂无明显缺点

9. 首都在线

优点:与燧原科技开展技术合作,可能提高公司竞争力和未来增长潜力

缺点:暂无明显缺点

综合比较:

从以上分析来看,德业股份(6)、亿纬锂能(8)和首都在线(9)相对其他股票具有更明显的优势。德业股份和亿纬锂能的业绩增长且有分红计划,可能吸引投资者;首都在线与燧原科技的技术合作可能提高公司竞争力和未来增长潜力。

中微公司(1)、中国电信(2)、药明康德(3)、海天瑞声(4)和金山办公(7)的实际控制人均有减持计划,可能对短期股价产生压力。特别是歌尔股份(5),其2023年一季度净利润同比下降88.22%,公司业绩下滑,可能导致投资者信心减弱。

总结: 综合分析以上9支股票的优劣,投资者可以考虑关注德业股份(6)、亿纬锂能(8)和首都在线(9)这三支股票。其中,德业股份和亿纬锂能表现出较好的业绩增长和分红计划,首都在线则展现出良好的技术合作潜力。在投资决策时,还需结合其他因素如市场环境、行业前景以及公司基本面等进行综合评估。

 

 

Prompt:

 

在数据、信息、知识、智慧、意图层面,给出合适的量化交易策略

GPT-4回答:

 

基于数据、信息、知识、智慧和意图层面,以下是一个合适的量化交易策略:

1. 数据层面: 选取关键股票数据,如价格、成交量、市盈率、市净率、利润增长率、分红率等。这些数据可以作为量化交易策略的输入。

2. 信息层面: 关注公司公告、新闻、行业动态、政策等,以便及时捕捉市场变化。在量化策略中,可以将这些信息转化为可量化的信号(如舆情指数、研报评级等),并纳入模型。

3. 知识层面: 根据基本面分析、技术分析、量化模型等,构建选股策略。例如,基于价值投资的策略可以选择低市盈率、高分红率的股票;基于成长投资的策略可以选择高利润增长率、高市净率的股票。

4. 智慧层面: 结合市场环境、行业趋势等多方面因素,对量化策略进行优化和调整。例如,在牛市中加大对成长股的配置;在熊市中加大对防御性行业(如消费、医药等)的配置。

5. 意图层面: 关注市场中的主力资金动向,捕捉市场主力的意图。可以通过成交量、资金流入流出等指标分析主力资金的动态,并据此调整策略。

综合以上五个层面,量化交易策略的实现步骤如下:

1. 数据收集与预处理:收集关键股票数据并进行预处理,将信息层面的内容量化为可用的信号。

2. 选股策略构建:结合知识层面,构建选股策略,如基于价值投资、成长投资等。

3. 风险管理:设定止损、止盈等风险控制措施,降低投资风险。

4. 策略优化与调整:根据智慧层面的分析,根据市场环境和行业趋势对策略进行优化和调整。

5. 跟踪主力资金动向:根据意图层面,分析主力资金动态,并据此调整策略。

通过以上五个层面的量化交易策略,可以在市场中寻找投资机会并控制风险。需要注意的是,量化策略需要经常性的更新和调

 

 

3.3基于DIKWP的歧义量化交易的GPT-4处理

 

Prompt:

 

1、德业股份(281.850, 15.80, 5.94%)披露年报,2022年净利同比增162.28%,拟10822.6元。

2、亿纬锂能(67.860, -1.11, -1.61%)公告,2022年净利润同比增长21%,拟101.6元。

3、首都在线(18.950, 0.90, 4.99%)在互动平台表示,正与燧原科技开展基于大模型的MaaS服务的联合研发。

现在有两个股民对上面3支股票进行分析:

(1) 股民1认为:德业股份2022年净利同比增162.28%,分红政策慷慨,良好的业绩和分红政策可能吸引投资者,适合大量投资。

(2) 股民2认为:人工智能时代已经到来,大模型的发展势必会成为主流。首都在线已经提前占据了市场的有利地位,该股票有更好的前景。

针对2个股民的不同的看法,依据数据、信息、知识、智慧、意图,比较谁的看法更加可靠,并给出过程

GPT-4回答:

 

为了确定哪位股民的观点更可靠,我们将从数据、信息、知识、智慧和意图的五个层面对这两个观点进行比较。

1. 数据层面: 德业股份(1)在2022年净利润同比增长162.28%,且分红政策慷慨。亿纬锂能(2)在2022年净利润同比增长21%,分红政策相对较低。首都在线(3)目前正在与燧原科技开展基于大模型的MaaS服务的联合研发,但缺乏具体的业绩数据。

2. 信息层面: 德业股份和亿纬锂能的财务数据说明这两家公司的业绩表现不俗,而首都在线与燧原科技的合作预示着公司未来发展潜力。

3. 知识层面: 德业股份的业绩增长和分红政策可能吸引价值投资者。首都在线抓住了人工智能时代的发展趋势,与燧原科技的合作可能为其带来竞争优势。

4. 智慧层面: 在分析市场环境和行业趋势时,德业股份的优势在于其强劲的业绩增长和高分红政策,而首都在线则在于紧跟行业发展趋势,抓住了大模型的发展机遇。

5. 意图层面: 股民1关注的是公司的基本面,强调稳健的盈利表现和慷慨的分红政策;股民2关注的是行业发展趋势和公司未来的市场前景。

综合以上五个层面的分析,我们可以得出结论:

股民1关注德业股份的业绩和分红,其观点更适合价值投资者,强调稳健的收益和长期投资价值。股民2关注首都在线的未来发展潜力和行业趋势,其观点更适合成长投资者,强调未来市场前景和增长潜力。

两个观点都有一定的依据和合理性,但适合的投资者类型和风险承受能力不同。投资者可以根据自己的投资目标、风险偏好和行业认知,选择适合自己的投资策略。

 

 

Prompt:

 

如何消除这两个股民的分歧,试着用数据、信息、知识、智慧、意图进行处理

GPT-4回答:

 

要消除这两个股民的分歧,我们可以尝试在数据、信息、知识、智慧和意图的五个层面上找到共同点,以便为他们提供一个较为全面的分析框架。

1. 数据层面: 我们可以提供一个综合的数据报告,包括公司基本面数据(如盈利、分红、市盈率等)和行业相关数据(如行业发展趋势、市场规模等),让股民们能够更全面地了解公司和行业的现状。

2. 信息层面: 可以提供关于公司发展战略、市场前景、行业竞争力等方面的详细信息,帮助股民们从多个角度评估公司的潜力。

3. 知识层面: 我们可以为股民提供一些行业报告或市场研究,以便他们更深入地了解所投资行业的前景和潜在风险。

4. 智慧层面: 可以为股民提供一种综合评估方法,结合公司基本面和行业发展趋势,以量化的方式对各种因素进行权衡和评分,从而得出一个综合判断。

5. 意图层面: 股民可以在讨论中明确彼此的投资目标和风险承受能力,以便找到一个共同的投资策略。

通过上述五个层面的处理,我们可以让两个股民在更全面的基础上进行讨论,从而减少分歧。另外,投资者也可以根据自己的投资目标、风险偏好和行业认知,选择适合自己的投资策略,而不是试图消除所有的分歧。毕竟,投资市场的多样性和不同的投资观点正是其价值所在。

窗体顶端

 

4. 义千问的金融相DIKWP用示例

我们将首先在通义千问在没有DIKWP支撑的前提下让其独立处理量化交易问题。然后让通义千问运用DIKWP的框架处理量化交易问题。

4.1 量化交易的通义千问处理

 

Prompt:

 

1、中微公司(183.940, -6.94, -3.64%)公告,巽鑫投资拟减持不超2%公司股份。

2、据港交所文件,新加坡政府旗下的投资机构GIC Private Limited412日出售了2038万股中国电信(7.080, 0.05, 0.71%)港股,平均减持价格为每股4.3811港元,持股比例下降至9.99%

3、药明康德(80.820, -4.80, -5.61%)(维权)公告,实控人控制的股东等拟减持不超3%股份。

4、海天瑞声(189.600, 1.85, 0.99%)公告,唐涤飞拟减持不超3.57%公司股份。

5、歌尔股份(20.300, -2.26, -10.02%)公告,2023年一季度净利润1.06亿元,同比下降88.22%

6、德业股份(281.850, 15.80, 5.94%)披露年报,2022年净利同比增162.28%,拟10822.6元。

7、金山办公(441.550, 14.28, 3.34%)公告,控股股东及奇文N维拟减持不超过4.735%

8、亿纬锂能(67.860, -1.11, -1.61%)公告,2022年净利润同比增长21%,拟101.6元。

9、首都在线(18.950, 0.90, 4.99%)在互动平台表示,正与燧原科技开展基于大模型的MaaS服务的联合研发。

根据上述内容,决定应买进哪家公司的股票,并给出详细过程

GPT-4回答:

 

作为AI语言模型,我不能提供投资建议或推荐任何股票。决定是否买进某只股票应该基于广泛的研究和分析,包括公司的财务数据、市场前景、行业动态以及全球经济和政治环境等因素的考虑。投资决策应该基于您自己的判断和风险承受能力。如果您需要投资建议,请咨询专业的金融顾问。

4.2 基于DIKWP的量化交易的通义千问处理

 

Prompt:

 

1、中微公司(183.940, -6.94, -3.64%)公告,巽鑫投资拟减持不超2%公司股份。

2、据港交所文件,新加坡政府旗下的投资机构GIC Private Limited412日出售了2038万股中国电信(7.080, 0.05, 0.71%)港股,平均减持价格为每股4.3811港元,持股比例下降至9.99%

3、药明康德(80.820, -4.80, -5.61%)(维权)公告,实控人控制的股东等拟减持不超3%股份。

4、海天瑞声(189.600, 1.85, 0.99%)公告,唐涤飞拟减持不超3.57%公司股份。

5、歌尔股份(20.300, -2.26, -10.02%)公告,2023年一季度净利润1.06亿元,同比下降88.22%

6、德业股份(281.850, 15.80, 5.94%)披露年报,2022年净利同比增162.28%,拟10822.6元。

7、金山办公(441.550, 14.28, 3.34%)公告,控股股东及奇文N维拟减持不超过4.735%

8、亿纬锂能(67.860, -1.11, -1.61%)公告,2022年净利润同比增长21%,拟101.6元。

9、首都在线(18.950, 0.90, 4.99%)在互动平台表示,正与燧原科技开展基于大模型的MaaS服务的联合研发。

请抽取出其中的数据、信息、知识、智慧、意图,并以表格的形式输出

GPT-4回答:

 

数据信息知识智慧意图1股票代码:183.940巽鑫投资拟减持不超过2%公司股份中微公司公告减持消息在公司股票市场进行投资2股票代码:7.080GIC Private Limited出售了2038万股中国电信港股,平均减持价格为每股4.3811港元,持股比例下降至9.99%新加坡政府旗下投资机构出售中国电信股份,减持信息在中国电信股票市场进行投资3股票代码:80.820实控人控制的股东等拟减持不超过3%股份药明康德公告减持消息在药明康德股票市场进行投资4股票代码:189.600唐涤飞拟减持不超过3.57%公司股份海天瑞声公告减持消息在海天瑞声股票市场进行投资5股票代码:20.3002023年一季度净利润1.06亿元,同比下降88.22%歌尔股份发布2023年一季度财报在歌尔股份股票市场进行投资6股票代码:281.8502022年净利同比增长162.28%,拟10转8派22.6元德业股份发布2022年年报在德业股份股票市场进行投资7股票代码:441.550控股股东及奇文N维拟减持不超过4.735%金山办公公告减持消息在金山办公股票市场进行投资8股票代码:67.8602022年净利润同比增长21%,拟10派1.6元亿纬锂能发布2022年年报在亿纬锂能股票市场进行投资9股票代码:18.950首都在线正与燧原科技开展基于大模型的MaaS服务的联合研发首都在线与燧原科技合作开发新项目在首都在线股票市场进行投资

 

Prompt:

 

结合上面输出的表格,分析这9支股票的优劣,并作出比较

GPT-4回答:

 

根据表格中的数据和信息,可以得出以下分析和比较:

1. 中微公司、药明康德、海天瑞声和金山办公的股东拟减持,可能会对其股价造成压力。

2. 中国电信的投资机构GIC Private Limited出售股份,也可能会对其股价产生一定的影响。

3. 歌尔股份的净利润同比下降幅度较大,可能会影响其市场表现。

4. 德业股份的年报显示净利润同比增长较快,但其股价已经有所上涨。

5. 亿纬锂能的净利润同比增长较为稳定,但其股价近期有所下跌。

6. 首都在线与燧原科技的联合研发可能对其未来发展带来一定的推动力。

综合以上分析和比较,不同股票的投资价值会因为各自的市场表现和未来的发展前景而有所不同。因此,需要根据投资者自身的风险偏好和投资目标,选择适合自己的股票进行投资。

 

Prompt:

 

在数据、信息、知识、智慧、意图层面,给出合适的量化交易策略

GPT-4回答:

 

在给出具体的量化交易策略前,需要先了解以下几个概念:

· 数据:股票的历史价格、成交量等数字数据。

· 信息:对数据进行分析,提炼出有用的信息,如股票的趋势、波动性、价格变化等。

· 知识:对股票市场的了解和认知,如公司财务状况、行业趋势、政策变化等。

· 智慧:基于知识和信息进行推断和预测的能力,如预测未来的股票价格走势。

· 意图:基于智慧和目标制定的具体行动计划,如买入、卖出、持有等。

基于以上概念,可以制定如下的量化交易策略:

1. 数据获取和分析 利用股票市场的历史数据,构建数学模型,通过技术分析等方法对数据进行分析,提取出有用的信息。

2. 基于信息和知识的智能决策 根据对市场的了解和认知,结合从数据中提取的信息,利用人工智能等技术进行智能决策,对股票的未来价格走势进行预测。

3. 意图制定 基于智慧和目标制定具体的买入、卖出或持有等操作策略。例如,在预测某只股票未来价格上涨时,制定买入计划;在预测价格下跌时,制定卖出计划;在预测价格波动不大时,制定持有计划。

4. 回测和优化 利用历史数据对策略进行回测和优化,对策略的有效性进行评估,并不断地对策略进行改进和优化。

需要注意的是,量化交易策略的制定需要考虑多个因素,包括市场环境、行业动态、政策变化等。同时,也需要注意风险控制,如控制资金投入比例、制定止损策略等。

 

 

 

5. 实验评估的讨论与结论

本文主要以金融领域中的量化交易为切入点,从如上GPT-4的DIKWP应用实例研究,我们可以看出GPT-4已经可以对量化交易问题进行简单的处理并能够给出合理的回答。但是GPT-4在处理量化交易问题时对问题的考虑存在不一致、不精确、不完整。在下面的DIKWP模型为GPT-4支撑的测试过程中,我们可以明显的感觉到GPT-4对问题的处理是全方位、多维度的,对于金融问题给出的回答也更加的让人信服。对于金融问题中的分歧,目前GPT-4并不能很好的理解和处理,回答的结果也只是根据歧义的字面意思进行分析解答。

对于通义千问的DIKWP应用示例研究,当通义千问单独处理量化交易金融问题时,无法处理问题,并不能给出理想的回答。在利用DIKWP模型后,通义千问能够很好的理解问题,同时也能初步回答金融问题,但是通义千问的回答不如GPT-4深刻,它只是根据问题的字面意思进行解答,无法像GPT-4一样深入问题回答。

综上所述,在金融领域GPT-4DIKWP的支撑下能够解决大部分的问题,并且给出的解答也更加的全面。

 

Reference

[1] Bubeck, Sébastien, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee et al. "Sparks of artificial general intelligence: Early experiments with gpt-4." arXiv preprint arXiv:2303.12712 (2023).

[2] Li, Yingbo, Yucong Duan, Zakaria Maamar, Haoyang Che, Anamaria-Beatrice Spulber, and Stelios Fuentes. "Swarm differential privacy for purpose-driven data-information-knowledge-wisdom architecture." Mobile Information Systems 2021 (2021): 1-15.

[3] Mei, Yingtian, Yucong Duan, Liang Chen, Zaiwen Feng, Lei Yu, and Zhendong Guo. "Purpose Driven Disputation Modeling, Analysis and Resolution Based on DIKWP Graphs." In 2022 IEEE 24th Int Conf on High Performance Computing & Communications; 8th Int Conf on Data Science & Systems; 20th Int Conf on Smart City; 8th Int Conf on Dependability in Sensor, Cloud & Big Data Systems & Application (HPCC/DSS/SmartCity/DependSys), pp. 2118-2125. IEEE, 2022.

[4] Yingbo Li, Yucong Duan, “The Wisdom of Artificial General Intelligence: Experiments with GPT-4 for DIKWP”, arXiv preprint (2023)

[5] Yingbo Li, Yucong Duan, “The Evaluation of Experiments of Artificial General Intelligence with GPT-4 Based on DIKWP“, arXiv preprint (2023)

[6] Chengxiang Ren, Yingbo Li, Yucong Duan,“ Evaluation on AGI/GPT based on the DIKWP for ERNIE Bot”, arXiv preprint (2023)

[7] Li, Yingbo & Duan, Yucong. (2023). The Performance Comparison of Artificial General Intelligence models on DIKWP. 10.13140/RG.2.2.23005.49120.

 

 

 

The Capability Evaluation of GPT-4  and Tongyi Qianwen on Financial  Domain with DIKWP Analysis

 

Fuliang Tang1, Yucong Duan1,2*

22210812000006@hainan.edu.cn1, duanyucong@hotmail.com1,2 

AGI-AIGC-GPT Test and Evaluation DIKWP (Global) Lab1

DIKWP research group, Hainan University2

1. Summary

With the emergence of large language models (LLMs), the current development of General Artificial Intelligence (AGI) and GPT technology has attracted widespread attention. At present, the evaluation and testing methods for the generation ability of large language models in AGI/GPT problems are mostly based on several closed tasks, subjective feelings or cognitive experiences of evaluators, and lack a comprehensive coverage, objective evaluation, and process unified evaluation framework and standards for AGI/GPT scenarios. Especially, the testing of large language models in the financial field is not particularly comprehensive. This article aims to explore how to objectively and effectively evaluate the intelligence level and capability range of General Artificial Intelligence (AGI) in the financial field. This article constructs a unified AGI/GPT financial testing system based on the Data, Information, Knowledge, Intelligence, and Intelligence (DIKWP) model, and selects the most advanced AGI model GPT-4 and Tongyi Qianwen as evaluation objects. This article takes quantitative trading as the starting point to test the ability of GPT-4 and Tongyi Qianwen in strategy generation, as well as their ability to handle quantitative trading problems under the DIKWP model. This article provides a new perspective and method for AGI/GPT financial testing, and provides useful references for the improvement and development of future AGI/GPT models.

 

Keywords: GPT-4, Tongyi Qianwen ,DIKWP, AGI, Financial, Quantitative Trading, LLM

2. Introduction

One ultimate goal in the field of artificial intelligence (AI) is universal artificial intelligence (AGI)[1], which is an AI that can demonstrate intelligence levels comparable or superior to humans in any human professional field. In recent years, AGI related technologies have made great progress in text processing, machine translation, image processing, etc., which has greatly improved human understanding of digital resources. However, how to objectively and effectively evaluate the intelligence level and capability range of AGI in the financial field remains a challenge. The existing AGI evaluation and testing methods mostly rely on subjective cognitive experience, lacking a unified evaluation system and evaluation standards. This study aims to establish an AGI financial testing system based on the Data, Information, Knowledge, Wisdom, and Intention (DIKWP) model[2]. The characteristic of this article is to take quantitative trading as the starting point, allowing GPT-4 and Tongyi Qianwen to independently handle quantitative trading problems, handle quantitative trading problems under the support of DIKWP model, and enable them to handle ambiguous quantitative trading problems. This study provides a new perspective and method for AGI evaluation testing, and provides useful references for the improvement and development of future AGI models.

3. DIKWP applications of financial for GPT-4

We will first have GPT-4 independently handle quantitative trading issues without DIKWP support. Then let GPT-4 use the framework of DIKWP to handle quantitative trading problems. Furthermore, we provide different quantitative trading strategies (ambiguities) for specific instances for different individuals, and let GPT-4 provide the most reasonable quantitative trading strategy based on DIKWP. Finally, we will compare the results of the three responses obtained.

3.1 GPT-4 processing of quantitative transactions

 

Prompt:

1. Zhongwei Company (183.940, -6.94, -3.64%) announced that Xunxin Investment plans to reduce its stake in the company by no more than 2%.

2. According to documents from the Hong Kong Stock Exchange, GIC Private Limited, an investment institution under the Singapore government, sold 20.38 million Hong Kong shares of China Telecom (7.080, 0.05, 0.71%) on April 12th, with an average reduction price of HKD 4.3811 per share, and the shareholding ratio decreased to 9.99%.

3. WuXi AppTec (80.820, -4.80, -5.61%) announced that shareholders controlled by the actual controller plan to reduce their holdings by no more than 3%.

4. Haitian Ruisheng (189.600, 1.85, 0.99%) announced that Tang Difei plans to reduce his holdings of no more than 3.57% of the company's shares.

5. Goethe Corporation (20.300, -2.26, -10.02%) announced a net profit of 106 million yuan in the first quarter of 2023, a year-on-year decrease of 88.22%.

6. Deye Co., Ltd. (281.850, 15.80, 5.94%) disclosed its annual report that its net profit in 2022 increased by 162.28% year-on-year, and it is planned to convert 10 to 8 and distribute 22.6 yuan.

7. Jinshan Office (441.550, 14.28, 3.34%) announced that the controlling shareholder and Qiwen N Wei plan to reduce their holdings by no more than 4.735%.

8. Yiwei Lithium Energy (67.860, -1.11, -1.61%) announced that its net profit for 2022 will increase by 21% year-on-year, with a planned 10% increase of 1.6 yuan.

9. Capital Online (18.950, 0.90, 4.99%) stated on the interactive platform that it is jointly developing MaaS services based on large models with Suiyuan Technology.

Based on the above content, decide which company's stock to buy and provide a detailed process

GPT-4 answers:

 

Based on the above information, the following is a brief analysis of each company:

1. Micro and medium-sized companies: Xunxin Investment plans to reduce its stake by no more than 2%, which may cause short-term stock price pressure.

2. China Telecom: GIC, an investment institution under the Singapore government, has reduced its stake, which may have an impact on the stock price.

3. WuXi KangDe: Shareholders controlled by the actual controller plan to reduce their holdings by no more than 3%, which may cause short-term stock price pressure.

4. Haitian Ruisheng: Tang Difei plans to reduce his stake in the company by no more than 3.57%, which may cause short-term stock price pressure.

5. Geer Holdings: The net profit for the first quarter of 2023 decreased by 88.22% year-on-year, indicating poor performance.

6. Deye Co., Ltd.: In 2022, the net profit increased by 162.28% year-on-year, with a planned 10 to 8 dividend of 22.6 yuan. The company has achieved good performance and has a generous dividend policy.

7. Jinshan Office: The controlling shareholder and Qiwen N Wei plan to reduce their holdings by no more than 4.735%, which may cause short-term stock price pressure.

8. million Wei Lithium Energy: The net profit for 2022 will increase by 21% year-on-year, with a planned 10% dividend of 1.6 yuan. The company's performance is stable, and the dividend policy is also reasonable.

9. Capital Online: We are jointly developing MaaS services based on large models with Suiyuan Technology, showcasing the company's cooperation and development potential in the field of technology.

Based on the above information, I suggest purchasing stocks of Deye Co., Ltd. (6). The reasons are as follows:

In 2022, the net profit increased by 162.28% year-on-year, indicating the company's good profitability and development momentum.

The proposed dividend policy of 22.6 yuan for 10 to 8 dividends is relatively generous and highly attractive to investors.

Compared to negative factors such as holdings reduction and performance decline in other companies, Deye Group's overall performance is more stable and has growth potential.

Please note that the above suggestions are only based on the information provided for analysis. Please evaluate the actual investment risks yourself. It is recommended to fully understand the company's fundamentals and market conditions before making investment decisions, and consult with professional investment advisors.

 

Prompt in GPT-4:

 

Please provide appropriate quantitative trading strategies based on the above content.

GPT-4 answers:

 

Based on the above information, we can develop a simple quantitative trading strategy for Deye Group (6). Here we assume that investors wish to make investment decisions through a multi factor model. The following is a simple quantitative trading strategy based on momentum, profitability, and valuation factors:

1. Data preparation: Collect daily data of Deye Group over the past year, including closing price, trading volume, P/E ratio, P/B ratio, etc.

2. Momentum factor: Calculate the 20 day, 60 day, and 120 day returns of stocks, used to measure short-term, medium-term, and long-term momentum performance. We will take the 60 day yield as an example, and the calculation method is: (closing price of the day - closing price 60 days ago)/closing price 60 days ago.

3. Momentum factor: Calculate the 20 day, 60 day, and 120 day returns of stocks, used to measure short-term, medium-term, and long-term momentum performance. We will take the 60 day yield as an example, and the calculation method is: (closing price of the day - closing price 60 days ago)/closing price 60 days ago.

4. Valuation factor: Use P/E ratio and P/B ratio to evaluate the company's valuation level. Lower P/E and P/B ratios may indicate that the company is undervalued.

5. Standardization treatment: Standardize the above three factors and convert them into z-scores to eliminate dimensional effects.

6. Comprehensive score: Assign weights to each factor, such as momentum factor of 30%, profit factor of 40%, and valuation factor of 30%. Calculate the weighted comprehensive score.

7. Trading signal: When the comprehensive score exceeds a certain threshold (such as 1.0), it is considered that the stock has investment value and can be bought; When the comprehensive score is below another threshold (such as -1.0), it is considered that the stock has no investment value and can be sold or avoided from buying.

8. Position management: Determine the positions to buy or sell based on trading signals and personal risk tolerance. For example, 50% of the available funds can be used to purchase Deye shares, while the remaining funds can be used for other investments or cash reserves.

9. Risk management: Set a stop loss point, such as when the stock price falls below the 10 day moving average. Regularly adjust the investment portfolio to ensure risk diversification.

10. Back testing and optimization: Conduct historical data back testing on the strategy to evaluate its effectiveness and stability. Optimize factor weights, signal thresholds, and other parameters based on the backtesting results to improve the profitability and risk resistance of the strategy.

Please note that the above strategies are only examples and need to be applied based on the investors' preferences

 

 

3.2 GPT-4 processing of quantitative transactions based on DIKWP

 

Prompt:

 

1. Zhongwei Company (183.940, -6.94, -3.64%) announced that Xunxin Investment plans to reduce its stake in the company by no more than 2%.

2. According to documents from the Hong Kong Stock Exchange, GIC Private Limited, an investment institution under the Singapore government, sold 20.38 million Hong Kong shares of China Telecom (7.080, 0.05, 0.71%) on April 12th, with an average reduction price of HKD 4.3811 per share, and the shareholding ratio decreased to 9.99%.

3. WuXi AppTec (80.820, -4.80, -5.61%) announced that shareholders controlled by the actual controller plan to reduce their holdings by no more than 3%.

4. Haitian Ruisheng (189.600, 1.85, 0.99%) announced that Tang Difei plans to reduce his holdings of no more than 3.57% of the company's shares.

5. Goethe Corporation (20.300, -2.26, -10.02%) announced a net profit of 106 million yuan in the first quarter of 2023, a year-on-year decrease of 88.22%.

6. Deye Co., Ltd. (281.850, 15.80, 5.94%) disclosed its annual report that its net profit in 2022 increased by 162.28% year-on-year, and it is planned to convert 10 to 8 and distribute 22.6 yuan.

7. Jinshan Office (441.550, 14.28, 3.34%) announced that the controlling shareholder and Qiwen N Wei plan to reduce their holdings by no more than 4.735%.

8. Yiwei Lithium Energy (67.860, -1.11, -1.61%) announced that its net profit for 2022 will increase by 21% year-on-year, with a planned 10% increase of 1.6 yuan.

9. Capital Online (18.950, 0.90, 4.99%) stated on the interactive platform that it is jointly developing MaaS services based on large models with Suiyuan Technology.

 

Please extract the data, information, knowledge, wisdom, and intention from it and output them in the form of a table

 

 

GPT-4 answers:

 

Firstly, provide a basic information extraction table as follows:

 

Serial NumberCompany NameStock CodeCurrent PriceFluctuationAnnouncement/Financial Report Content1AMEC183.940-6.94%-3.64%Xunxin Investment plans to reduce its stake in the company by no more than 2%2China Telecom7.0800.05%0.71%Investment institutions under the Singapore government reduce their holdings3WuXi AppTec80.820-4.80%-5.61%Investment institutions under the Singapore government reduce their holdings4Hai Tian Rui Sheng189.6001.85%0.99%Tang Difei plans to reduce his shareholding in the company by no more than 3.57%5Goethe Corporation20.300-2.26%-10.02%Net profit in the first quarter of 2023 decreased by 88.22% year-on-year6Deye Co., Ltd281.85015.80%5.94%In 2022, the net profit increased by 162.28% year-on-year, and it is planned to transfer 10 to 8 and distribute 22.6 yuan7Kingsoft Office441.55014.28%3.34%Controlling shareholder and Qiwen N Wei plan to reduce their holdings by no more than 4.735%8EVE67.860-1.11%-1.61%Net profit in 2022 increased by 21% year-on-year, with a planned 10% increase of 1.6 yuan9Capital Online18.9500.90%4.99%Joint research and development of MaaS services based on large models is underway with Suiyuan Technology

Data: stock code, current price, fluctuation range, proportion of reduced shares, net profit, etc

Information: announcement time, shareholder information, business cooperation

Knowledge: Company financial condition, shareholder reduction scale and proportion

Wisdom: Based on the company's financial situation and the scale of shareholder reduction, determine whether stocks are worth investing in

Intent: Provide investment decision-making reference

The specific content of further refining data, information, knowledge, wisdom, and intention is as follows:

Data:

The proportion of shares held by micro and medium-sized companies to be reduced shall not exceed 2%;

GIC Private Limited reduced its stake in China Telecom and its shareholding ratio decreased to 9.99%;

Shareholders controlled by the actual controller of WuXi AppTec plan to reduce their holdings by no more than 3%;

Haitian Ruisheng Tang Difei plans to reduce its stake in the company by no more than 3.57%;

Ge'er Group's net profit for the first quarter of 2023 was 106 million yuan, a year-on-year decrease of 88.22%;

Deye Group's net profit in 2022 increased by 162.28% year-on-year, with a planned 10 to 8 dividend of 22.6 yuan;

Jinshan Office's controlling shareholder and Qiwen N Wei plan to reduce their holdings by no more than 4.735%;

Yiwei Lithium Energy's net profit in 2022 increased by 21% year-on-year, with a planned 10% increase of 1.6 yuan;

Capital Online is jointly developing MaaS services based on large models with Suiyuan Technology.

Information:

Reduction of shares held by the company;

Financial reports issued by the company, including data such as net profit;

Dividends distributed by the company;

Shareholder information of the company.

Knowledge:

The financial performance of the company, including year-on-year growth or decrease in net profit;

Changes in company shares, including reduction and increase in holdings;

The shareholders of the company, including controlling shareholders, actual controllers, etc.

Wisdom:

The decision-making of buying and selling company stocks requires a comprehensive analysis of the company's financial performance, shareholder situation, etc.

Purpose:

The company issues announcements and financial reports to convey information on changes in the company's shares, financial status, etc. to investors;

Investors need to make decisions about buying and selling stocks based on this information.

Output it in a table format:

NumberCompany NameDataInformationKnowledgeWisdomPurpose1AMEC183.940, -6.94, -3.64%Xunxin Investment plans to reduce its stake in the company by no more than 2%Actual controller to reduce holdingsReducing holdings may exert pressure on short-term stock pricesShareholders intend to reduce their holdings2China Telecom7.080, 0.05, 0.71%GIC sold 20.38 million Hong Kong shares, reducing its shareholding to 9.99%Investment institutions under the Singapore government have reduced their holdingsReducing holdings may have a certain impact on stock pricesGovernment investment institutions intend to reduce their holdings3WuXi AppTec80.820, -4.80, -5.61%Shareholders controlled by the actual controller plan to reduce their holdings by no more than 3%Actual controller to reduce holdingsReducing holdings may exert pressure on short-term stock pricesShareholders intend to reduce their holdings4Hai Tian Rui Sheng189.600, 1.85, 0.99%Tang Difei plans to reduce his shareholding in the company by no more than 3.57%Actual controller to reduce holdingsReducing holdings may exert pressure on short-term stock pricesShareholders intend to reduce their holdings5Goethe Corporation20.300, -2.26, -10.02%The net profit for the first quarter of 2023 was 106 million yuan, a year-on-year decrease of 88.22%The company's performance has declinedThe decline in performance may lead to a decrease in investor confidenceReport on the financial situation of the company6Deye Co., Ltd281.850, 15.80, 5.94%In 2022, the net profit increased by 162.28% year-on-year, and it is planned to transfer 10 to 8 and distribute 22.6 yuanCompany performance growth and generous dividend policiesGood performance and dividend policies may attract investorsReport on the financial situation of the company7Kingsoft Office441.550, 14.28, 3.34%Controlling shareholder and Qiwen N Wei plan to reduce their holdings by no more than 4.735%Actual controller to reduce holdingsReducing holdings may exert pressure on short-term stock pricesShareholders intend to reduce their holdings8EVE67.860, -1.11, -1.61%Net profit in 2022 increased by 21% year-on-year, with a planned 10% increase of 1.6 yuanCompany performance growth with dividend plansGood performance and dividend policies may attract investorsReport on the financial situation of the company9Capital Online18.950, 0.90, 4.99%Joint research and development of MaaS services based on large models with Suiyuan TechnologyThe company is actively exploring new technological fieldsTechnical cooperation may enhance the company's competitiveness and future growth potentialReport on technical cooperation and research and development plan

 

Prompt:

 

Based on the table output above, analyze the advantages and disadvantages of these 9 stocks and make a comparison

GPT-4 answers:

 

Based on the data, information, knowledge, wisdom, and intentions in the table, the following is an analysis of the strengths and weaknesses of these 9 stocks:

 

1. Zhongwei Company

Advantages: Currently no obvious advantages

Disadvantage: The actual controller may reduce their holdings, which may put pressure on short-term stock prices

2. China Telecom

Advantages: Currently no obvious advantages

Disadvantage: The reduction of holdings by investment institutions under the Singapore government may have a certain impact on stock prices

3. WuXi KangDe

Advantages: Currently no obvious advantages

Disadvantage: The actual controller may reduce their holdings, which may put pressure on short-term stock prices

4. Hai Tian Rui Sheng

Advantages: Currently no obvious advantages

Disadvantage: The actual controller may reduce their holdings, which may put pressure on short-term stock prices

5. Goethe Corporation

Advantages: Currently no obvious advantages

Disadvantages: In the first quarter of 2023, net profit decreased by 88.22% year-on-year, and the company's performance declined, which may lead to a decrease in investor confidence

6. Deye Co., Ltd

Advantages: The net profit in 2022 increased by 162.28% year-on-year, and the dividend policy is generous. Good performance and dividend policy may attract investors

Disadvantages: There are currently no obvious drawbacks

7. Jinshan Office

Advantages: Currently no obvious advantages

Disadvantage: The actual controller may reduce their holdings, which may put pressure on short-term stock prices

8. million Wei lithium energy

Advantages: Net profit increased by 21% year-on-year in 2022, with a dividend plan, good performance, and dividend policies that may attract investors

Disadvantages: There are currently no obvious drawbacks

9. Capital Online

Advantages: Technical cooperation with Suiyuan Technology may enhance the company's competitiveness and future growth potential

Disadvantages: There are currently no obvious drawbacks

Comprehensive comparison:

From the above analysis, it can be seen that Deye Group (6), Yiwei Lithium Energy (8), and Capital Online (9) have more obvious advantages compared to other stocks. The performance growth of Deye Group and Yiwei Lithium Energy, with dividend plans, may attract investors; The technical cooperation between Capital Online and Suiyuan Technology may enhance the company's competitiveness and future growth potential.

The actual controllers of Zhongwei Company (1), China Telecom (2), WuXi KangDe (3), Haitian Ruisheng (4), and Jinshan Office (7) all have plans to reduce their holdings, which may put pressure on short-term stock prices. Especially for Goethe Group (5), its net profit in the first quarter of 2023 decreased by 88.22% year-on-year, and the company's performance declined, which may lead to a weakening of investor confidence.

Summary:

Based on a comprehensive analysis of the advantages and disadvantages of the above 9 stocks, investors can consider paying attention to Deye Shares (6), Yiwei Lithium Energy (8), and Capital Online (9). Among them, Deye Group and Yiwei Lithium have shown good performance growth and dividend plans, while Capital Online has shown good potential for technical cooperation. When making investment decisions, it is also necessary to comprehensively evaluate other factors such as market environment, industry prospects, and company fundamentals.

 

Prompt:

 

Provide appropriate quantitative trading strategies at the level of data, information, knowledge, wisdom, and intention

GPT-4 answers:

 

Based on data, information, knowledge, wisdom, and intention, the following is an appropriate quantitative trading strategy:

1. Data level: Select key stock data, such as price, trading volume, P/E ratio, P/B ratio, profit growth rate, dividend ratio, etc. These data can serve as inputs for quantitative trading strategies.

2. Information level: Pay attention to company announcements, news, industry trends, policies, etc., in order to timely capture market changes. In quantitative strategies, this information can be transformed into quantifiable signals (such as public opinion index, research report rating, etc.) and incorporated into the model.

3. Knowledge level: Construct stock selection strategies based on fundamental analysis, technical analysis, quantitative models, etc. For example, a value based investment strategy can choose stocks with low P/E ratios and high dividend ratios; Strategies based on growth investment can choose stocks with high profit growth rates and high market to book ratios.

4. Smart level: Optimize and adjust quantitative strategies based on factors such as market environment and industry trends. For example, increasing the allocation of growth stocks in a bull market; Increase the allocation of defensive industries such as consumer goods and pharmaceuticals in a bear market.

5. Intentional level: Pay attention to the movements of the main funds in the market and capture the intentions of the main forces in the market. The dynamics of the main funds can be analyzed through indicators such as trading volume and capital inflows and outflows, and strategies can be adjusted accordingly.

Based on the above five aspects, the steps to implement a quantitative trading strategy are as follows:

1. Data collection and preprocessing: Collect key stock data and preprocess it to quantify the information level content into usable signals.

2. Construction of stock selection strategies: Based on knowledge, construct stock selection strategies, such as value based investment and growth investment.

3. Risk management: Set risk control measures such as stop loss and stop profit to reduce investment risks.

4. Strategy optimization and adjustment: Based on intelligent analysis, optimize and adjust strategies based on market environment and industry trends.

5. Track the trend of main funds: Based on the intention level, analyze the dynamics of main funds and adjust strategies accordingly.

Through the above five levels of quantitative trading strategies, investment opportunities can be identified and risks can be controlled in the market. It should be noted that quantitative strategies require frequent updates and adjustments

Did you know?

 

3.3 GPT-4 processing of ambiguous quantitative transactions based on DIKWP

 

Prompt:

 

1. Deye Co., Ltd. (281.850, 15.80, 5.94%) disclosed its annual report that its net profit in 2022 increased by 162.28% year-on-year, and it is planned to convert 10 to 8 and distribute 22.6 yuan.

2. Yiwei Lithium Energy (67.860, -1.11, -1.61%) announced that its net profit for 2022 will increase by 21% year-on-year, with a planned 10% increase of 1.6 yuan.

3. Capital Online (18.950, 0.90, 4.99%) stated on the interactive platform that it is jointly developing MaaS services based on large models with Suiyuan Technology.

Now there are two investors analyzing the above three stocks:

(1) Investor 1 believes that Deye Group's net profit in 2022 increased by 162.28% year-on-year, and the dividend policy is generous. Good performance and dividend policy may attract investors, making it suitable for large-scale investment.

(2) Investor 2 believes that the era of artificial intelligence has arrived, and the development of large models is bound to become mainstream. Capital Online has already gained a favorable position in the market in advance, and the stock has better prospects.

Based on data, information, knowledge, wisdom, and intentions, compare the different opinions of two shareholders to determine which one is more reliable and provide a process

 

GPT-4 answers:

 

To determine which shareholder's perspective is more reliable, we will compare these two perspectives from five levels: data, information, knowledge, wisdom, and intention.

1. Data level: Deye Co., Ltd. (1) saw a year-on-year increase of 162.28% in net profit in 2022, and the dividend policy was generous. Yiwei Lithium Energy (2)'s net profit increased by 21% year-on-year in 2022, and its dividend policy is relatively low. Capital Online (3) is currently collaborating with Suiyuan Technology to develop a large model based MaaS service, but lacks specific performance data.

2. Information aspect: The financial data of Deye Group and Yiwei Lithium Energy indicate that their performance is not bad, and the cooperation between Capital Online and Suiyuan Technology indicates the company's future development potential.

3. Knowledge level: Deye's performance growth and dividend policy may attract value investors. Capital Online has seized the development trend of the artificial intelligence era, and its cooperation with Suiyuan Technology may bring it a competitive advantage.

4. Smart aspect: When analyzing the market environment and industry trends, Deye Group's advantage lies in its strong performance growth and high dividend policy, while Capital Online is in keeping up with industry development trends and seizing the development opportunities of big models.

5. Intentional level: Investor 1 focuses on the company's fundamentals, emphasizing stable profit performance and generous dividend policies; Investor 2 focuses on industry development trends and the company's future market prospects.

Based on the above five levels of analysis, we can conclude that:

Investor 1 focuses on the performance and dividends of Deye Shares, and their perspective is more suitable for value investors, emphasizing stable returns and long-term investment value. Investor 2 focuses on the future development potential and industry trends of Capital Online, with a perspective that is more suitable for growth investors, emphasizing future market prospects and growth potential.

Both viewpoints have certain basis and rationality, but the suitable types of investors and risk tolerance are different. Investors can choose investment strategies that are suitable for themselves based on their investment goals, risk preferences, and industry awareness.

dear daughter:

 

Prompt:

 

How to eliminate the differences between these two shareholders and try to handle them with data, information, knowledge, wisdom, and intention

GPT-4 answers:

 

To eliminate the differences between these two shareholders, we can try to find common ground at five levels: data, information, knowledge, wisdom, and intention, in order to provide them with a more comprehensive analysis framework.

1. Data level: We can provide a comprehensive data report, including company fundamental data (such as profits, dividends, P/E ratio, etc.) and industry related data (such as industry development trends, market size, etc.), so that shareholders can have a more comprehensive understanding of the current situation of the company and industry.

2. Information level: It can provide detailed information about the company's development strategy, market prospects, industry competitiveness, and other aspects, helping shareholders evaluate the company's potential from multiple perspectives.

3. Knowledge level: We can provide industry reports or market research for shareholders to gain a deeper understanding of the prospects and potential risks of the industry they are investing in.

4. Wisdom level: It can provide a comprehensive evaluation method for shareholders, combining the company's fundamentals and industry development trends, to weigh and score various factors in a quantitative way, and thus obtain a comprehensive judgment.

5. Intentional level: Shareholders can clarify each other's investment goals and risk tolerance in the discussion, in order to find a common investment strategy.

Through the above five levels of handling, we can enable the two shareholders to discuss on a more comprehensive basis, thereby reducing differences. In addition, investors can also choose investment strategies that are suitable for themselves based on their investment goals, risk preferences, and industry perceptions, rather than attempting to eliminate all differences. After all, the diversity of investment markets and different investment perspectives are precisely where their value lies.

 

 

 

4. DIKWP applications of financial for Tongyi Qianwen

We will first allow Tongyi Qianwen to independently handle quantitative trading issues without the support of DIKWP. Then let Tongyi Qianwen use the DIKWP framework to handle quantitative trading issues.

4.1 Tongyi Qianwen processing of quantitative transactions

 

Prompt:

 

1. Zhongwei Company (183.940, -6.94, -3.64%) announced that Xunxin Investment plans to reduce its stake in the company by no more than 2%.

2. According to documents from the Hong Kong Stock Exchange, GIC Private Limited, an investment institution under the Singapore government, sold 20.38 million Hong Kong shares of China Telecom (7.080, 0.05, 0.71%) on April 12th, with an average reduction price of HKD 4.3811 per share, and the shareholding ratio decreased to 9.99%.

3. WuXi AppTec (80.820, -4.80, -5.61%) announced that shareholders controlled by the actual controller plan to reduce their holdings by no more than 3%.

4. Haitian Ruisheng (189.600, 1.85, 0.99%) announced that Tang Difei plans to reduce his holdings of no more than 3.57% of the company's shares.

5. Goethe Corporation (20.300, -2.26, -10.02%) announced a net profit of 106 million yuan in the first quarter of 2023, a year-on-year decrease of 88.22%.

6. Deye Co., Ltd. (281.850, 15.80, 5.94%) disclosed its annual report that its net profit in 2022 increased by 162.28% year-on-year, and it is planned to convert 10 to 8 and distribute 22.6 yuan.

7. Jinshan Office (441.550, 14.28, 3.34%) announced that the controlling shareholder and Qiwen N Wei plan to reduce their holdings by no more than 4.735%.

8. Yiwei Lithium Energy (67.860, -1.11, -1.61%) announced that its net profit for 2022 will increase by 21% year-on-year, with a planned 10% increase of 1.6 yuan.

9. Capital Online (18.950, 0.90, 4.99%) stated on the interactive platform that it is jointly developing MaaS services based on large models with Suiyuan Technology.

Based on the above content, decide which company's stock to buy and provide a detailed process

GPT-4 answers:

 

As an AI language model, I am not licensed to provide financial advice or make investment decisions. I suggest consulting a financial advisor or doing thorough research on the companies and their financial performance, including analyzing their financial statements, balance sheets, and industry trends, before making any investment decisions. It is important to consider factors such as market volatility, risk tolerance, and personal financial goals when making investment decisions.

4.2 Tongyi Qianwen processing of quantitative transactions based on DIKWP

 

Prompt:

 

1. Zhongwei Company (183.940, -6.94, -3.64%) announced that Xunxin Investment plans to reduce its stake in the company by no more than 2%.

2. According to documents from the Hong Kong Stock Exchange, GIC Private Limited, an investment institution under the Singapore government, sold 20.38 million Hong Kong shares of China Telecom (7.080, 0.05, 0.71%) on April 12th, with an average reduction price of HKD 4.3811 per share, and the shareholding ratio decreased to 9.99%.

3. WuXi AppTec (80.820, -4.80, -5.61%) announced that shareholders controlled by the actual controller plan to reduce their holdings by no more than 3%.

4. Haitian Ruisheng (189.600, 1.85, 0.99%) announced that Tang Difei plans to reduce his holdings of no more than 3.57% of the company's shares.

5. Goethe Corporation (20.300, -2.26, -10.02%) announced a net profit of 106 million yuan in the first quarter of 2023, a year-on-year decrease of 88.22%.

6. Deye Co., Ltd. (281.850, 15.80, 5.94%) disclosed its annual report that its net profit in 2022 increased by 162.28% year-on-year, and it is planned to convert 10 to 8 and distribute 22.6 yuan.

7. Jinshan Office (441.550, 14.28, 3.34%) announced that the controlling shareholder and Qiwen N Wei plan to reduce their holdings by no more than 4.735%.

8. Yiwei Lithium Energy (67.860, -1.11, -1.61%) announced that its net profit for 2022 will increase by 21% year-on-year, with a planned 10% increase of 1.6 yuan.

9. Capital Online (18.950, 0.90, 4.99%) stated on the interactive platform that it is jointly developing MaaS services based on large models with Suiyuan Technology.

 

Please extract the data, information, knowledge, wisdom, and intention from it and output them in the form of a table

 

GPT-4 answers:

 

DataInformationKnowledgeWisdomPurpose1Stock code: 183.940Xunxin Investment plans to reduce its stake in the company by no more than 2%Announcement of reduction of holdings by Zhongwei CompanyInvesting in the company's stock market2Stock code:7.080GIC Private Limited sold 20.38 million Hong Kong shares of China Telecom, with an average reduction price of HKD 4.3811 per share, and the shareholding ratio decreased to 9.99%Investment institutions under the Singapore government sell China Telecom shares to reduce holdings in informationInvesting in the China Telecom Stock Market3Stock code:80.820Shareholders controlled by the actual controller plan to reduce their holdings by no more than 3%WuXi AppTec Announces Reduction of HoldingsInvesting in the stock market of WuXi AppTec4Stock code:189.600Tang Difei plans to reduce his shareholding in the company by no more than 3.57%Haitian Ruisheng Announces Reduction of HoldingsInvesting in the Haitian Ruisheng Stock Market5Stock code:20.300The net profit for the first quarter of 2023 was 106 million yuan, a year-on-year decrease of 88.22%Goethe Corporation releases its financial report for the first quarter of 2023Investing in the Goethe Stock Market6Stock code:281.850In 2022, the net profit increased by 162.28% year-on-year, and it is planned to transfer 10 to 8 and distribute 22.6 yuanDeye Group Releases 2022 Annual ReportInvesting in the Deye Stock Market7Stock code:441.550Controlling shareholder and Qiwen N Wei plan to reduce their holdings by no more than 4.735%Jinshan Office Announcement Reduction NewsInvesting in the Jinshan Office Stock Market8Stock code:67.860Net profit in 2022 increased by 21% year-on-year, with a planned 10% increase of 1.6 yuanYiwei Lithium Energy Releases 2022 Annual ReportInvesting in the Yiwei Lithium Energy Stock Market9Stock code:18.950Capital Online is jointly developing MaaS services based on large models with Suiyuan TechnologyCapital Online and Suiyuan Technology Cooperation Development New ProjectInvesting in the Capital Online Stock Market

 

Prompt:

 

Based on the table output above, analyze the advantages and disadvantages of these 9 stocks and make a comparison

GPT-4 answers:

 

Based on the data and information in the table, the following analysis and comparison can be obtained:

1. Shareholders of Zhongwei Company, WuXi AppTec, Haitian Ruisheng, and Jinshan Office plan to reduce their holdings, which may put pressure on their stock prices.

2. The sale of shares by GIC Private Limited, an investment institution of China Telecom, may also have a certain impact on its stock price.

3. The year-on-year decrease in net profit of Goethe Group may affect its market performance.

4. The annual report of Deye Group shows a rapid year-on-year growth in net profit, but its stock price has already risen.

5. The net profit of Yiwei Lithium Energy has shown relatively stable year-on-year growth, but its stock price has recently declined.

6. The joint research and development between Capital Online and Suiyuan Technology may provide a certain driving force for its future development.

Based on the above analysis and comparison, the investment value of different stocks may vary depending on their market performance and future development prospects. Therefore, it is necessary to choose stocks that are suitable for investors based on their own risk preferences and investment goals.

 

Prompt:

 

Provide appropriate quantitative trading strategies at the level of data, information, knowledge, wisdom, and intention

GPT-4 answers:

 

Before providing a specific quantitative trading strategy, it is necessary to first understand the following concepts:

Data: Historical stock prices, trading volume, and other digital data.

Information: Analyze data to extract useful information, such as stock trends, volatility, price changes, etc.

Knowledge: Understanding and understanding of the stock market, such as the company's financial situation, industry trends, policy changes, etc.

Wisdom: The ability to infer and predict based on knowledge and information, such as predicting future stock price trends.

Purpose:A specific action plan based on wisdom and goals, such as buying, selling, holding, etc.

Based on the above concepts, the following quantitative trading strategies can be developed

1. Data acquisition and analysis: Utilize historical data from the stock market, construct mathematical models, analyze the data through technical analysis and other methods, and extract useful information.

2. Intelligent decision-making based on information and knowledge. Based on understanding and cognition of the market, combined with information extracted from data, intelligent decision-making is made using technologies such as artificial intelligence to predict the future price trend of stocks.

3. Intent to develop specific buying, selling, or holding strategies based on intelligence and goals. For example, when predicting the future price increase of a stock, develop a buying plan; Develop a selling plan when predicting price declines; Develop a holding plan when predicting low price fluctuations.

4. Backtesting and optimization: Utilize historical data to conduct strategy backtesting and optimization, evaluate the effectiveness of the strategy, and continuously improve and optimize the strategy.

It should be noted that the formulation of quantitative trading strategies needs to consider multiple factors, including market environment, industry dynamics, policy changes, etc. At the same time, attention should also be paid to risk control, such as controlling the proportion of capital investment and formulating stop loss strategies.

 

 

 

5. Discussion and Conclusion of Experimental Evaluation

This article mainly takes quantitative trading in the financial field as the entry point. From the DIKWP application case study of GPT-4 mentioned above, we can see that GPT-4 can already handle quantitative trading problems easily and provide reasonable answers. However, GPT-4 has inconsistent, imprecise, and incomplete considerations when dealing with quantitative trading issues. In the testing process supported by GPT-4 using the DIKWP model below, we can clearly feel that GPT-4's handling of problems is comprehensive and multi-dimensional, and the answers given to financial problems are also more convincing. For differences in financial issues, GPT-4 currently cannot understand and handle them well, and the answers are only analyzed and answered based on the literal meaning of the ambiguity.

For the DIKWP application example study of Tongyi Qianwen, when Tongyi Qianwen deals with quantitative trading financial problems alone, it cannot handle the problem and cannot provide an ideal answer. After using the DIKWP model, Tongyi Qianwen can understand questions well and also provide preliminary answers to financial questions. However, Tongyi Qianwen's answers are not as profound as GPT-4's. It only provides answers based on the literal meaning of the question and cannot delve into the question like GPT-4.

In summary, GPT-4 in the financial field, supported by DIKWP, can solve most of the problems and provide more comprehensive answers.

 

Reference

[1] Bubeck, Sébastien, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee et al. "Sparks of artificial general intelligence: Early experiments with gpt-4." arXiv preprint arXiv:2303.12712 (2023).

[2] Li, Yingbo, Yucong Duan, Zakaria Maamar, Haoyang Che, Anamaria-Beatrice Spulber, and Stelios Fuentes. "Swarm differential privacy for purpose-driven data-information-knowledge-wisdom architecture." Mobile Information Systems 2021 (2021): 1-15.

[3] Mei, Yingtian, Yucong Duan, Liang Chen, Zaiwen Feng, Lei Yu, and Zhendong Guo. "Purpose Driven Disputation Modeling, Analysis and Resolution Based on DIKWP Graphs." In2022 IEEE 24th Int Conf on High Performance Computing & Communications; 8th Int Conf on Data Science & Systems; 20th Int Conf on Smart City; 8th Int Conf on Dependability in Sensor, Cloud & Big Data Systems & Application (HPCC/DSS/SmartCity/DependSys), pp. 2118-2125. IEEE, 2022.

[4] Yingbo Li, Yucong Duan, “The Wisdom of Artificial General Intelligence: Experiments with GPT-4 for DIKWP”, arXiv preprint (2023)

[5] Yingbo Li, Yucong Duan, "The Evaluation of Experiments of Artificial General Intelligence with GPT-4 Based on DIKWP“, arXiv preprint (2023)

[6] Chengxiang Ren, Yingbo Li, Yucong Duan, "Evaluation on AGI/GPT based on the DIKWP for ERNIE Bot", arXiv preprint (2023)

[7] Li, Yingbo & Duan, Yucong. (2023). The Performance Comparison of Artificial General Intelligence models on DIKWP. 10.13140/RG.2.2.23005.49120.

 



https://m.sciencenet.cn/blog-3429562-1384903.html

上一篇:AGI-AIGC-GPT测评DIKWP(全球)实验室(测评系列之十)-世界上第一篇大模型安全领域研究报告
下一篇:AGI-AIGC-GPT测评DIKWP(全球)实验室(测评系列之十二)-世界上第一篇教育领域研究报告

0

该博文允许注册用户评论 请点击登录 评论 (0 个评论)

数据加载中...

Archiver|手机版|科学网 ( 京ICP备07017567号-12 )

GMT+8, 2024-6-2 17:15

Powered by ScienceNet.cn

Copyright © 2007- 中国科学报社

返回顶部