Model-based Digital Business Reporting - Charlie

http://xbrl.squarespace.com/digital-financial-reporting/

这是Charlie近今年一直在关心研究的基本成功,XBRL 模型。

Model-based Digital Business/FinancialReporting

 

This page summarizes information related todigital financial reporting. Be sure to check out the wiki for digitalfinancial reporting. You can follow this topic by following the category"digital financial reporting" on my blog.

 

Modelling Business Information Using XBRL.A resource for accountants, internal auditors, external auditors, financialanalysts, and other business professionals who want to create, review, audit,or extract information from model-based digital business/financial reports.

 

Table of Contents

1.        Introduction

2.        Getting Started

3.        Understanding Important KeyIdeas, Concepts, Terminology

4.        Understanding theMultidimensional Model

5.        Logical Model Report Elements

6.        Information Models(Metapatterns)

7.        Domain Partion AggregationModels

8.        Overview of Business Use Cases

9.        Business Use Cases

10.    Integrity Models

11.    Flow Models

12.    Comprehensive Example

13.    SEC XBRL Financial FilingModel/Reference Implementation

14.    Specific or Special ModelingConsiderations

15.    Semantic Models

16.    Verifying Digital BusinessReports Using Automated Validation

17.    Analysis and Comparison ofDigital Business Information

18.    Remodelled US GAAP Taxonomy andExemplars

19.    APPENDIX: Analysis of 1474 SECXBRL Filings

20.    (Blank)

21.    APPENDIX: Report ElementProperties

22.    APPENDIX: Top XBRL TechnicalSyntax Related Modeling Tips

23.    APPENDIX: Restricting DataTypes

24.    APPENDIX: UnderstandingCalculation Inconsistencies

25.    APPENDIX: Why the SEC May Move to Inline XBRL

Steps To Mastering Digital FinancialReporting

 

NOTE: The information below is nowsummarized and included within the one document above. The separate PDFs areidentical to the combined PDF above.

 

The following is a summary of steps tomastering model-based digital financial reporting. Eventually the number ofsteps will be reduced once software better supports the needs of businessusers. No business user should need to understand the details of the XBRLtechnical syntax. The technical details will eventually be hidden from mostusers.  Regretfully, we are not at thispoint yet. So, understanding the correct approach to using XBRL in digitalfinancial reporting and the things you need to do to create quality digitalfinancial reports is a little more challenging today. This information isparticularly helpful to consultants and others who are in a "train thetrainer" type role.

Step 1. Understanding the Fundamentals(REQUIRED). The first step in understanding how to use XBRL for digitalfinancial reporting is to understand some fundamental things about what XBRLis, what it is not, etc.  This video HowXBRL Works is a very good step in that direction. This blog post expands on thevideo in a little more detail.

Step 2. Get Some Background(OPTIONAL).  This is an optional step,but it can be useful in filling in gaps in understanding you might have. Thebest accumulation of background material is the book I wrote, XBRL for Dummies,with the purpose of summarizing that background.

Step 3. XBRL Technical Primer (OPTIONAL).Chapter 4 of the XBRL for Dummies book (about 25 pages) has a very good primerfor whose who want background in the XBRL technical syntax.  This step can help you fill in gaps inunderstanding. Although not required for your journey, it can be very helpfulfor those who truly want to master XBRL. Additionally, if you want tounderstand why XBRL is what it is, read chapter 2 of XBRL for Dummies, Takingto Heart the Essential Concepts of XBRL. This will help you understand thereasoning behind decisions made and sometimes conflicting objectives.

Step 4. Hello World Example (OPTIONAL).Chapter 15 of XBRL for Dummies walks you through a few "hello world"type examples to help you get your feet wet using XBRL. The walk through is alittle technical, but only enough to expose you to important new ideas.

Step 5. Understand the MultidimensionalModel (REQUIRED). The key ideas of the multidimensional model are important tounderstand before you begin your journey in earnest.  This brief overview is all you really need, acouple of pages worth.

Step 6. Overview of a Logical Model ofDigital Financial Reporting (REQUIRED). These are summaries of report elementsand relations between report elements.  Alogical model or sometimes called a conceptual model takes something morecomplex and articulates it in a manner which is easier to understand.  For example, the logical model of anelectronic spreadsheet (workbooks, spreadsheets, columns, rows, cells) followsthe logical model of a paper spreadsheet. The logical model used for digitalfinancial reporting helps hide the XBRL technical syntax from business users.

Step 7. Report Elements and Report ElementRelations (REQUIRED). A digital business report contains report elements.Report elements can be put into categories. Report elements can be related toother report elements. This document goes into more details about reportelements and report element relations. (OPTIONAL, for more details, thisdocument covers the details of report element properties: This providesadditional information as to report element properties.

Step 8. Information Model MetapatternsDetails (REQUIRED). Information models/metapatterns. Computer systemsarchitects and engineers leverage patterns to make things easier for users oftheir systems. Financial reporting information has patterns. Everything infinancial reporting can be created using one or more of these 8 metapatternswhich are the building blocks of all financial reporting information.

Step 9. Domain Partition Aggregation ModelsDetails (REQUIRED). Domain partition aggregation models. The members within adomain may, or may not, aggregate or add up. This document explains the relation models between a set of memberswhich makes up a domain.

Step 10. Integrity Models Details(REQUIRED). Integrity Models. Relations exist within a table, but relationsalso exist between tables. Integrity models express the semantic relationsbetween the components of one table and the components of another table.

Step 11. Flow Models Detail (REQUIRED).Flow Models. Flow is the notion of relations between networks and/or tables forthe purpose of ordering or sequencing the components of a digital financialreport.

Step 12. Business Use Cases (REQUIRED).Business Use Cases. This helps you understand how to model information inspecific situations. It helps you see the subtle differences between differentmodeling situations. This provides sample files, documentation, and otherinformation for about 30 business use cases and helps you understand eachbusiness use case and compare/contrast different business use cases.

Step 13. Tying Business Use Cases Together(REQUIRED). Comprehensive Example. While the metapatterns and business usecases are helpful isolated examples/samples; these metapatterns and businessuse cases have to work together within larger reports.  The comprehensive example pulls each of thebusiness use cases together to examine the integrity between different businessuse cases.

Step 14. (SEC focused) Tying EverythingTogether with a Reference/Model Implementation (REQUIRED). Referenceimplementation of SEC XBRL financial filing. This reference implementation orprototype or model provides an example of an SEC XBRL financial filing. The keybenefit which this provides is to see how the individual pieces of the filingrelate to one another. The business rules, articulated using XBRL Formula, isthe best place to see that each piece is intact and each piece relates properlyto other pieces.


  • 0
    点赞
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
### 回答1: model-free和model-based是机器学习中的两种不同方法。 model-free指的是一种无模型的学习方法,它不需要事先建立一个模型来描述数据的生成过程,而是直接从数据中学习出一个策略或者函数来解决问题。例如,强化学习中的Q-learning算法就是一种model-free的方法。 model-based则是一种基于模型的学习方法,它需要先建立一个模型来描述数据的生成过程,然后再从模型中学习出一个策略或者函数来解决问题。例如,基于贝叶斯网络的分类器就是一种model-based的方法。 两种方法各有优缺点,选择哪种方法取决于具体的问题和数据特征。 ### 回答2: Model-free 和 model-based 是机器学习中常见的两种方法。它们的主要区别在于建模的方式和预测结果的依据不同。 Model-free 是一种不依赖于事先建立数学模型的方法。它只关心输入数据和输出结果之间的关系,而不需要理解其背后的机制。这种方法比较简单易用,可以处理各种类型的数据,例如声音、图像、文本等,而且在实时预测和决策时能够快速准确。常见的 model-free 算法包括神经网络、支持向量机、决策树、随机森林以及强化学习等。它们通过对数据进行训练和学习,从而得到模型的预测结果。 Model-based 是一种基于数学模型的方法。它建立数学模型来描述数据之间的关系,然后从模型中推导出预测结果。这种方法需要先对数据进行建模和参数估计,能够提高预测精度和正确率,而且可以进行更加严谨的理论分析。但是它对数据的类型和数据的质量要求比较高,需要在建模前对数据进行预处理和特征工程。常见的 model-based 算法包括线性回归、逻辑回归、贝叶斯网络、高斯过程以及概率图模型等。它们通过通过定义模型和求解参数,对数据进行建模和预测。 在实际应用中,通常需要综合考虑 model-free 和 model-based 的方法。根据具体的任务需求和数据特性,选择适合的方法。例如,当数据量大、类型多元化、噪声较多时,可以采用 model-free 的方法。当数据质量高、预测结果需要较高的精度和可解释性时,可以采用 model-based 的方法。在训练和测试时,可以采用交叉验证、调参和集成等技术来提高预测效果。 ### 回答3: Model-basedmodel-free 是强化学习(Reinforcement Learning)中的两个重要概念。这两种方法都是解决在特定环境下的最优策略问题。但两种方法的思路略有不同,分别适用于不同的场景。 Model-based 是一种通过先建立一个对当前环境的精确模型,再依据这个模型进行最优策略的计算的方法。这种方法的优点是可以准确无误地预测出采取某个决策之后的结果,通过模拟可能的动作序列来计算最优策略,并且能够分析出策略的稳定期望回报。但是,这种方法的缺点是需要预测模型,这需要对系统环境有足够的先验知识,同时模型复杂度高,需要大量的时间和计算资源。 相比之下,model-free 方法不需要事先预测环境模型,直接维护一个价值函数(value function)或者一个策略(policy),利用对当前状态的观测来训练价值函数或者策略,从而最终获得最优策略。model-free 方法的优点是不需要精确的环境模型,易于实现和解释,而且可以迭代地优化产生的策略。然而,相对于 model-based 方法,model-free 方法需要采样的数据更多,并且需要更长时间来确定最优策略,有可能产生收敛到局部最小值或者产生不稳定性的情况。 总的来说,两种方法各具优劣,选择哪种方法需要根据具体问题及环境来分类。在某些环境中可以采取自适应机制,按照当前情况选择 model-basedmodel-free 方法。对于新问题一般首先进行 model-free 试验,然后使用 model-based 去完善分析。因此,两种方法不是相互排斥的,而是在特定场景下相互补充和协同的。

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值