ACCT systems I exam 3
packaged in blocks and chained together, thus, this technology eventually became popularized as blockchain.
The internal data structure of transactions in the system is:
Data Visualization
The process of presenting information graphically
Processing Power:
The processing power required to obtain information valuable to the company could be enormous or even impossible.
more test data or real-world data.
The trained model is applied to
1. Middleman 2. Delay 3. Service fee
Three important concepts involved in the traditional transactions:
-Client nodes -Peer nodes -Ordering nodes
Three types of nodes when discussing hyperleder
-Distributed and decentralized. The data are distributed and synchronized among all the participants in the network. -Consensus. All parties will be aware of transactions that take place on the network and agree to the transactions being written to the blockchain. -Immutability. Once transactions are confirmed on the blockchain, they are tamperproof and cannot be altered.
To be characterized as blockchain, there must be a few components that are met:
**Traditional Systems: -System is centralized. -Requires a middleman to approve and record transactions. -Only one copy of the ledger **Blockchain Systems: -System is decentralized, distributed ledger -No middleman needed, multiple copies -When a new transaction occurs, all nodes are in sync. -Information cannot be added or deleted without the knowledge of the entire network. -A write-once, read-many system.
Traditional vs. Blockchain system:
means to electronically communicate business information to facilitate business reporting of financial and nonfinancial data to users.
XBRL serves as a
facilitate business reporting of financial and non-financial data to users.
XBRL serves as a means to electronically communicate business information and
eXtensible Business Reporting Language
XBRL stands for
programmable in Ethereum. This flexibility drew the attention of corporations and government agencies.
"Smart contracts" are
Supervised Learning
-Output is a known set of values. -Neural network seeks to predict the output by using the input dataset. -The data includes pairs of input and output.
Semi-Supervised Learning
-Some data are labeled, but some labels can be incorrect or missing. -The computer discerns values for the incorrect and missing outputs: Active Learning: queries the user to discover the right label/output.
Reinforcement Learning
-The machine/model learns by trial-and-error. -The program acts and learns by feedback from those actions as it works toward a desired goal. -The goal is dynamic. It can change over time based on circumstance.
Unsupervised Learning
-Uses unstructured data rather than labeled data -No input-output pairs -The goal usually involves: Clustering or dimension reduction; Identify outliers in the input dataset -Different parameters can result in different results.
$3 trillion in value per year in just a subset of industries impacted.
A study from McKinsey Global Institute estimates that Big Data could generate up to
$5.8 trillion annual value across all business sectors.
AI can create up to
Correct classifications/total values: (True pos+True neg)/ (true pos+false pos+false neg+True neg)
Accuracy ratio:
1. reporting independent 2. system independent 3. permits consolidation 4. provides flexibility (allowing extensible, flexible, multi-national solutions that can exchange the data required by internal finance, accountants, and creditors)
Advantages of XBRL GL
Descriptive Analysis:
Analysis performed that characterizes, summarizes and organizes past performance.
Diagnostic Analysis:
Analysis performed to investigate the underlying cause of a phenomenon
Predictive Analysis:
Analysis performed to provide foresight by identifying patterns in historical data.
prescriptive analysis
Analysis performed which identifies the best possible options given constraints or changing conditions
1. Ask the Question 2. Master the data 3. Perform the analysis 4. Share the story
The AMPS Model stands for:
"Your Data Won't Speak Unless You Ask It the Right Questions."
The AMPS Model: Ask the Question
1. Data Accessibility - can we get the needed data to answer the question posed? 2. Data Reliability - is the data clean? 3. Data Integrity - is the data accurate, valid and consistent over time? 4. Data Type - is the data structured? is the data internal? are there privacy concerns with the data?
The AMPS Model: Master the Data: Data questions include:
thinking about the data analytics process.
The AMPS model is a means of
1. Which product is most profitable at stores in Missouri? 2. Is it more profitable to produce an item in the United States or in Mexico (or Indonesia)? 3. Why are our costs increasing in the West but decreasing in the East? 4. What is the probability that our audit client will go bankrupt or need to restate its financial statements?
The AMPS model potential questions include:
addressed with data and that lead to a better decision making.
The AMPS model starts with asking questions that can be
-Continuous Audit Becomes Possible (Due to the distributed nature and real-time information retrieval of blockchain technology, it is possible for auditors to conduct a continuous audit over time.) -Opportunities for Future Roles in CPA (Smart contracts is one of the key areas: -Understand the business rules embedded in the smart contracts and the related libraries. -Verify the appropriateness of tax code and regulations in the smart contracts. )
The Impact of Blockchain on Audit and Assurance
1. What Happened? - Descriptive Analysis 2. Why Did it Happen? - Diagnostic Analysis 3. Will it Happen in the Future? - Predictive Analysis 4. What Should We Do, Based on What We Expect Will Happen? - Prescriptive Analysis
The Type of Question Asked Leads to the Analysis Performed
various uses, including reporting on the firm's web site, filing to regulators (SEC, IRS, etc.) and providing information to other interested parties such as financial analysts, loan officers and investors.
The XBRL database is available for
-Thinking logically -Acting rationally -Visual perception -Speech recognition -Language translation
The ability of computers to perform tasks that associated with human intelligence, such as:
analyze and assess data in a way that helps business makers make decisions. We call this process data analytics.
The abundance of data availability gives new opportunities to:
salaries of the data analytics scientists and the cost of the technology to prepare and analyze the data.
The cost to scrub the data includes the
volume, velocity, veracity and variety
The four Vs
represent the defining features of Big Data
The four Vs are often used to
descriptive, diagnostic, predictive and prescriptive analytics.
The four types of analytics performed include
the speed and accuracy of business reporting.
XBRL greatly enhances
the XML language, a standard for Internet communication between businesses
XBRL is based on
optimize predictions.
A cleaned and well-defined training data set is used to
cognitive technologies
Artificial Intelligence applications are also called
intelligence exhibited by machines rather than humans.
Artificial Intelligence:
-Reduces the time and effort involved in accessing data by -Works well with standard audit and risk analytic tests often run against datasets in specific accounts or groups of accounts (such as inventory or accounts receivable or sales revenue transactions). -Allows software vendors (such as ACL Inc.) to produce data extraction programs for given enterprise systems to help facilitate fraud detection and prevention and risk management. -Facilitates testing of the full population of transactions, rather than just a small sample. -Connects/interacts well with XBRL GL Standards (to be introduced in Chapter 10).
Audit Data Standards provide the following benefits:
it must be scrubbed from extraneous data and noise.
Before data can be analyzed and be useful
-first cryptocurrency. -eliminates the ability to double spend. -Anonymous peer-to-peer transactions, no middleman involved. -Public blockchain: anyone can join or leave at any time. -Validation through proof of work and rewards as an economic incentive via a resource intensive computation called mining. -Immutable history of transactions. -Distributed ledger. -One block is added to the blockchain approximately every 10 minutes. -The First-Mover: the first blockchain application in production. -Transaction fees differ by block size -Mining reduces by 50 percent every 4 years -SHA-256 -Mining reward is 12.5 Bitcoin -Higher transaction fee -Unspent transaction output system (spending cash and receiving change)
Bitcoin:
-Supply chain: EX: transporting temperature-related goods (ice cream, milk) -Loyalty Program: Ex: customer's frequent flier miles -Auto Industry Ex: a whole process from customizing order to delivering car
Blockchain uses cases
accumulate and analyze data that might be helpful to the firm's strategic initiatives.
Business intelligence uses computer-based techniques to
spend less time looking for evidence, which will allow more time for presenting their findings and making judgments.
By using data analytics, auditors are able to
- Illegal, fraudulent, or unauthorized transactions can reside in the blockchain - The records residing in the blockchain network may not provide sufficient information for auditors. - Complicated protocols involved in the distributed ledger system - An auditor needs to understand how the business is implemented and how the business uses blockchain technology in its end-to-end business applications. - Different blockchain platforms with different protocols and terms make it difficult to apply traditional audit approaches to blockchain technology. - Acquiring the skill sets to understand smart contracts is crucial to audit blockchain use cases.
Challenges on Audit and Assurance:
self-learning algorithms that allow computers to examine connections and notice patterns without human intervention
Cognitive technologies employ...
storage and processing
Companies generally face two important limiting factors in their business systems when dealing with Big Data:
--True positive (TP): the correct prediction of positive class: spam email was predicted correctly --True negative (TN): the correct prediction of negative class: correct prediction of email that is not spam. --False positive (FP): the incorrect prediction of positive class: an email was predicted as spam but actually was not spam --False negative (FN): the incorrect prediction of negative class: an email was predicted as not spam but actually was spam.
Confusion Matrix:
- Permissioned blockchain - Allows several organizations to participate in. - Administrators establish the access rights and permissions (private channels) for each participant. - Executed only on a limited set of trusted nodes. - Permit more complicated enterprise behaviors. / \ Hyperledger Corda
Consortium Blockchain:
**Corda: Known identity Selective endorsement Asset Smart contracts Access control rules Appropriate privacy **Public Blockchain: Anonymity Proof of work Cryptocurrency Fixed business rules No defined roles Public transaction information
Corda v. public blockchain
R3
Corda, developed by
-Uses smart contracts for its business rules. -Only relevant parties can join the network. -Only related parties will be informed for each transactions. -The administrators define and restrict each user's access rights. -Only a restricted set of trusted nodes execute the consensus protocol.
Corda:
-Blockchain protocols are lacking in areas such as speed, confidentiality, and governance requirements. -Most enterprises are opting to start with permissioned or private blockchain networks, which require a method to govern who is eligible to participate in the network. -Challenges in integrating private blockchain network with existing enterprise solutions
Current Challenges with Adopting Blockchain Technology
1. Understand the data 2. Select the data visualization tool: •Excel •Tableau •Power BI •Others 3. Develop and present the visualization: •Create or reinforce knowledge •Choose the right chart
Data Visualization Process:
sharing the story and turning data into information.
Data Visualization is one way of
services such as testing for fraudulent transactions and automating compliance-monitoring activities (e.g., filing financial reports with the SEC or IRS).
Data analytics also expands auditors' capabilities in
technologies, systems, practices, methodologies, databases, and applications used to analyze diverse business data to help organizations make sound and timely business decisions.
Data analytics often involves the
future of audit.
Data analytics plays a very critical role in the
50 percent and 90 percent of their time cleaning data for analysis.
Data analytics professionals estimate that they spend between
accounting and auditing in particular.
Data analytics will increasingly be critical for business in general and for
data warehouse to meet a specific need.
Data marts represent a slice of data from the
used to find patterns in stock prices to assist technical financial stock market analysts, or in commodities or currency trading
Data mining is often
statistical relationships and some of them represent spurious correlations. Data mining must be coupled with common sense to interpret the statistical relationships found.
Data mining will only find
relevant information to decision makers
Data visualization presents:
separate from the operational database.
Data warehouses are kept
facilitate decision making such as those often used in managerial accounting
Data warehouses are often designed to
operational systems to provide necessary insight, particularly in the case of customer relationship management (CRM) and supply chain management (SCM) systems
Data warehouses do work together with
main repository of the firm's historical data, or in other words, its corporate memory and will often serve as an archive of past firm performance
Data warehouses often serve as the
he operating databases of the firm to support decision making across a number of functions in the firm.
Data warehouses serve as a repository of information that is separate from
a type of Machine Learning.
Deep Learning is
-form of machine learning that involves complex, multilayer (deep) neural networks. -Has more than two nonoutput layers -Provides more power to solve sophisticated problems but also creates model complexity.
Deep learning is a
-Did we make a profit last year? -How much did we pay in federal taxes last year? -How long have the existing accounts receivable been past due?
Descriptive Analysis addresses questions like:
Digital Dashboard
Designed to track the firm process or performance indicators or metrics to monitor critical performance.
-Why did advertising expense increase, but sales fall? -Why did we experience an unfavorable labor rate variance last year? -Why did overall tax increase even though net income did not?
Diagnostic Analysis addresses questions:
easily accessible to executives.
Digital dashboards tracks critical firm performance in a way that is
access standard reports (i.e. 10-K going to the SEC or the corporate tax return going to the IRS) or specialized reports (i.e. accessing only specific data for a financial analyst, etc.)
Each interested XBRL user can either
-cryptocurrency -One block is added every 10-12 seconds -Transaction fees differ by computational complexity -Mining of Ether occurs at a constant rate -Ethash algorithm reduces the advantage of ASIC (application-specific integrated circuit) in mining -Mining reward is 3 Ether -Lower transaction fee -Account is debited/credited -The Ethereum virtual machine is used to run the Ethereum smart contracts.
Ethereum:
-Machine learning -Neural networks -Robotic process automation (RPA) -Bots -Natural language processing
Examples of cognitive technologies include:
-What is the chance the company will go bankrupt? -What is our expected sales and income next year? -Can we predict if the financial statements will be misstated? -Will the borrower pay us back the loan we've granted her?
Predictive Analysis addresses questions like:
Linux Foundation
Hyperledger was created by
-Smart contracts, configurable consensus, and member management services. -Private channels -Membership identify service and the access control list (ACL) provides an additional layer of role-based access control. -Blockchain log file: keep track the provenance of transactions -Store current state of blockchain to facilitate the speed of verification (SQL command) -Assets are added, updated, and transferred with chaincode.
Hyperledger:
this cost of cleaning and formatting the data could be alleviated
If both the provider and the user (e.g., a company and its external auditor) of the data had the same data standards for their data:
eliminate the need for intermediaries in trustless, online, peer-to-peer digital currency transactions.
In 2009, Nakamoto used a distributed ledger system through resource intensive mining to
pull together logic and business rules into contracts represented in code called "smart contracts" through Ethereal.
In 2014, blockchain 2.0 emerged as a more robust and sophisticated technology to:
-The transactions are done without any middleman involved. -Much faster transaction time (minutes vs days). -Lower service fee.
In a blockchain system:
internal processes, improving productivity, utilization, and growth.
In addition to producing more value externally, studies show that data analytics affects
scan the environment—that is, by scanning social media to identify potential risks and opportunities to the firm.
In financial accounting, data analytics may be used to
data warehouses can run data queries without slowing down the performance of the company's operational systems.
Information in the warehouse can be stored safely for extended periods of time and
the computer's ability to learn from experience rather than specific instructions.
Machine learning involves
a type of AI;
Machine learning is
Storage:
Many companies choose to use a cloud platform to lower the cost of data storage.
-Classification: seeks to assign labels, dividing the input into output groups, such as:- Yes or No- Spam or Not Spam -Regression: seeks to predict real numbers, such as:- The price of a house- The revenue in next quarter
Most machine learning applications are designed to perform either
mathematical models that convert inputs to outputs/predictions, can be nested together
Neural networks are
-Recruit auditors with enough understanding of blockchain technology/ IT background and offer training in smart contracts -Recruit software and IT background people and do the onsite audit training. -Form an audit team that combines the expertise of IT, taxation, law, accounting, and auditing.
New Approaches for Future Audit Tasks:
Web Crawler
One way that firms may gather business intelligence is by use of a
variance reports, trend reports, variance analysis reports, and reports that show actual performance are compared to budgeted information
Output might include
identify future opportunities and risks.
Patterns discovered from historical data enable businesses to
-What is the level of sales needed to breakeven? -How can revenues to maximized if there is a trade war with China? -Should the company lease or buy its headquarters office? -Should the company make its own products or outsource production to another company?
Prescriptive Analysis addresses questions like:
- Permissioned blockchain/ Enterprise blockchain - Requires permission to join the network. - Transaction data and validation are restricted. - Not expose internal information to the public.
Private Blockchain
1. Gather information from a variety of sources 2. Analyze the data to discern patterns and trends from that information to gain understanding and meaning 3. Make decisions based on the information gained
Process of Business Intelligence
-The administrator identities who creating blocks are known and reputable. -The rest of the network can vote for admin removal in case of malicious behavior found in the network.
Proof of Authority:
-A set of validators who propose the next block lock up an amount of their cryptocurrency as a deposit to ensure honest behavior. -It reduced computer costs and centralization risks.
Proof of Stake:
-All miners compete to create the next block to be committed to the blockchain. This is done by solving a complex mathematical problem. -It requires the miners use computer power so it can prevent attacks as it requires tons of computer powers to overcome
Proof of work:
-Permissionless blockchain -No access restrictions in viewing or participation -Offers economic reward for the computational proof of work in mining. / \ Bitcoin ether
Public Blockchain
time consuming.
Reformatting, cleansing, and consolidating large volumes of data from multiple sources and platforms can be especially
Natural Language Processing (NLP) and Natural Language Understanding (NLU): - Understand text - Extract semantic meaning - Discern sentiment from natural language Robotic process automation (RPA): A tool that can perform high-volume repetitive accounting tasks. Machine Learning in Audit and Assurance: - GL.ai (PWC) - replicate the thinking and decision making - Deloitte's machine - review contracts, identify key terms - Helix GLAD (EY): detect anomalies in large database - CLARA (KPMG): potential risks - BDO neural networks - manage info in multiple languages globally
Types of AI applications in accounting:
unstructured and unprocessed data, such as comments in social media, emails, global positioning system (GPS) measurements, etc.
Variety refers to
act that the data comes in at quick speeds or in real time, such as streaming videos and news feeds.
Velocity refers to the
quality of the data including extent of cleanliness (without errors or data integrity issues), reliability and representationally faithful.
Veracity refers to the
Machine learning models learn from training cases or data. The models then make predictions for new products or data. Those new products then generate new data, which improves the ability to make predictions.
Virtuous cycle of machine learning:
massive amount of data involved
Volume refers to the
1. The most current standardized XBRL taxonomy. 2. Reliable underlying financial and non-financial data in XBRL tagging. 3. Accurate and complete XBRL tagging. 4. Accurate, complete, and received on timely basis XBRL generate reports.
We believe XBRL Assurance should include
1. Enable multiple parties that do not fully trust each other to collaborate with a shared source of truth. 2. Accelerate transaction settlement and verification by eliminating intermediaries. 3. Help cut costs and resources that would be spent on manual verification (help auditors collecting and evaluating evidence to support transactions).
When is blockchain useful:
discover various patterns, investigate anomalies, forecast future behavior, and so forth.
With a wealth of data on their hands, companies are empowered by using data analytics to
anything that is found in a chart of accounts, journal entries or historical transactions, financial and non-financial
XBRL GL allows the representation of
XBRL Global Ledger Taxonomy
XBRL GL is also known as
Data Warehouse
a collection of information gathered from an assortment of external and operational (i.e., internal) databases to facilitate reporting for decision making and business analysis.
Artificial neural networks
are the engines of machine learning.
Business Intelligence
computer-based technique for accumulating and analyzing data from databases and data warehouses to support managerial decision making
XBRL instance documents
contain the actual dollar amounts or the details of each of the elements within the firm's XBRL database.
Big data
datasets that are too large and complex for businesses' existing systems to handle using their traditional capabilities to capture, store, manage and analyze these data sets
Data Analytics
defined as the science of examining raw data, removing excess noise and organizing the data with the purpose of drawing conclusions for decision making.
The XBRL taxonomy
defines and describes each key data element (e.g., total assets, accounts payable, net income, etc.)
Feed-forward neural networks
information moves in one direction.
Data mining
one technique used to analyze data for business intelligence purposes. A process using sophisticated statistical techniques to extract and analyze data from large databases to discern patterns and trends that were not previously known.
correct positive predictions/# of positive predictions: true pos/ (false pos+true pos)
percision ratio:
correct positive predictions/true positive values: true positive/ (true pos+ False neg)
recall ratio
Audit Data Standards (ADS)
set of standards for data files and fields typically needed to support an external audit in a given financial business process area.
XBRL style sheets
take the instance documents and add presentation elements to make them readable by humans.
Recurrent neural networks
the connections between neurons include loops.