The process of using Data Mining in an Audit

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

The growth of world has changed many things which also drastically changed the business environment. Perfect companies with excellent operation and management skills also have the chances of business failure and insolvency if they do not have a proper tool for data storage and retrieval. During the early studies, techniques such as multiple discriminant analysis, univariate and logit have been used as a data retrieval method but those methods use history samples. However, with the growth of information technology, data mining techniques rise up where it helps to store and retrieve valuable financial data of a company. (Jie Sun, 2006) The main purpose of data mining is to identify and mine out valid patterns of data which can provide knowledgeable information. (Hian Chye Koh, 2004) The world became lack of confidence in audit judgements when there was a lot of audit fraud and failure, for an example the failure of Enron. Since than it became an important role for auditors to take responsibility of audit purpose and improve the audit structure. (R.Jayalakshmy,) Data mining became highly important in business environment especially in audit field. Auditors have the responsibility to estimate the possibilities of management frauds during audit process. They can achieve this by using data mining (DM) techniques where progress is identified by discovering either through automatic or manual way. Several law enforcement and investigation units have used data mining techniques effectively to identify fraudulent activities. (Efstathios Kirkos, 2007) American institute of Certified Public Accountants (AICA, 1999) has acknowledged data mining as one of the top ten rank technologies. Nevertheless, Institute of Internal Auditors has listed data mining as one of the top rank research priorities. (Hian Chye Koh, 2004) There are several data mining techniques which can be used in auditing. The main techniques which are widely used are neural networks, decision tree and Bayesian belief networks. (Efstathios Kirkos, 2007) Each of these techniques has its own advantages and disadvantages. Choosing the most appropriate techniques is what this research going to discover.

Survey of Literature

Data mining techniques plays an important role in today's era as it helps to retrieve and collect valuable data from huge data warehouse. The collaboration of accounting and information technology leads to the term e-audit. Application of data mining techniques in auditing influenced business and organizations in many ways. Research and studies have been done on this issue explaining the actual need of data mining in auditing. There are several issues extracted from these research journals and articles, those are; (1) Types of data mining methods, (2) Preventing audit fraud using data mining, (3) The need of continuous auditing, (4) How data mining change the perspectives of auditor and audit.

Data mining and types of data mining methods

Mu-Yen Chen and Yin-Kuan Du, (2008) mentioned that data mining is a process of gathering data from large database or data warehouse which can provide meaningful information. Data mining is also known as knowledge discovery in databases (KDD) because it provides useful information for decision making process. There are several iterative steps in data mining process concerning following steps:

Application domain identification

In this step, some research will be carried forward to understand well the application domain and the appropriate knowledge domain. This is done to identify the actual goal of the KDD.

Target dataset

Selecting and focusing on a dataset which is relevant to the analysis and task. Those datasets are retrieved from the relevant databases.

Data pre-processing

Data will be reprocessed according to the need where unwanted data will be removed and cleaned. The main objective here is to get the accurate data.

Data mining

This is the main process in data mining where different techniques and rules are applied such as association rule, neural networks, decision tree and Bayesian theorem in order to get meaningful information.

Knowledge extraction

Once any of the techniques or rules has been applied, the information will provide some useful knowledge relating to the analysis. This method also rechecks and resolves any differences with previously believed knowledge.

Knowledge application

The knowledge gained from previous process will be directly applied to the domain for further action and development.

Knowledge evaluation

The accuracy of the mined knowledge will be improved from time to time.

Data mining techniques have been applied in many business fields. It has been successfully applied to many financial domain especially audit field.

Together with the data mining steps mentioned, there are also further data mining algorithms and process. Classification process where data item is classified into several predefined categories. Regression process will be used in data mining where data will be mapped into a real-value prediction variable. Clustering process will map the data into a cluster based on similarity metrics or probability density model. Use association rule to describe the relationship between different attributes. Summarizations will provide a short description for the subset of data. Dependency modelling will describe the significant dependencies among those variables and sequence analysis will be used to model sequential patterns by generating the sequence or extract and report deviations over time.

In the study conducted by Mu-Yen and Yin-Kuan, they describe the data mining process on how it will generally extract financial data and manipulate it according to the needs. They also discuss on the accuracy of this manipulated information. It is widely known that audit is a process of handling large sets of financial data and accounting statements.

The research takes an example of Taiwan Stock Exchange Corporation (TSEC). The huge sets of data will undergo cleaning and pre-processing in order to remove discrepancies and inconsistencies. The main aim in this phase is to identify and select suitable indicators including financial and non-financial ratios. The next phase will load this indicators discovery prediction rule will be ready to be applied in data mining clustering. In the modelling phase financial statement will be collected for data mining processing where association or back propagation rules will be used to analyze the data sets. Mu-Yen and Yin-Kuan (2008).

In data mining method, a pre-processing method know as attribute oriented induction (AOI) and information gain (IG) will be used in order to get efficient decision with more meaningful data. Jie and Hui(2008). There are three main types of data mining method that can be used in audit; those are neural network, decision tree and Bayesian.

Data mining techniques such as neural networks, decision tree and Bayesian networks can be used to for predictive modelling. Researchers found that data mining techniques can predict better compare to traditional statistic methods when multifaceted nonlinear and interaction relationship exist in the collection of data. Hian and Chan (2004)

Neural network

Neural networks are capable of handling inconsistent data and are most suitable to be applied when no algorithmic solution is needed. There are several types of neural networks such as self-organizing maps (SOM) and back propagation. SOM have only input and output layer whereas back propagation has additional one several hidden layers. Advantage of neural networks is that it is capable of handling inconsistent data. It also does not require algorithm solution and does not make assumptions about attributes independence. Efstathios, Charalambos, and Yannis (2007).

Neural network plays an important role in audit by recognizing patterns in the data. It is very useful when the relationship between the dependent and independent data is unknown. Neural network consist of nodes which function similar like neurons in human brain. The input layer consists of independent variable while the output layer consists of dependent variable. There are hidden nodes between the input and output nodes. Each node in neural network will perform a computation and transformation in order to combine each input and generate an output. Neural network can be applied in any non-linear function relationship since hidden nodes or layers can represent latent combinations of unobservable variables. The application of neural network in auditing provides an interesting illustration of data mining in cross-sectional business applications. Hian and Chan (2004).

Decision tree

Decision tree is more accurate compared to both neural and Bayesian method since it is not based on hypothesis and the result is fast and accurate. Decision tree is viewed as a tree shaped structure where the structure is formed from collection of data. The leaf attribute represent a class while the non-leaf node represent testing of an attribute value. Jie and Hui(2008).

There are several splitting algorithms used in decision tree. Those famous algorithms are the chi-square Automatic Interaction Detection which uses the chi-square statistic, the Classification and Regression Trees (CART) which uses index of diversity and ID3 which uses an entropy-based measure. There are possibilities where the division or splitting of the decision tree produces a large tree. This can be avoided by using tree pruning where it removes the splitting nodes in a way that it does not affect the accuracy rate. Tree pruning will be used in decision tree in order to remove false outliers so that it will not significantly affect the model's accuracy rate. The advantages of using decision tree in auditing are that it will reflect a more significant way of presenting gained knowledge and it is easy to extract by using the IF-THEN rules. Efstathios, Charalambos, and Yannis (2007).

Decision tree is mainly used for predicting or classifying data. It works by dividing findings into mutually exclusive and exhaustive subgroups. The process of division is based on the dependent and independent variables. The division process keeps continuing till either no further meaningful splitting or no significant differences with the dependent variable. Hian and Chan (2004).


Another type of data mining technique is Bayesian theorem. There are two main types of Bayesian methods Naïve Bayesian and Bayesian belief networks. Naïve Bayesian makes the class state as independence assumption whereas Bayesian belief network (BBN) permit for illustration of dependencies among subsets of attributes. Naïve Bayesian gives the best accuracy rates but in many cases this assumption is not valid because dependencies exist between most of the attributes. Bayesian belief network allow dependent assumption between the attributes. It is a directed acyclic graph where each node represents an attribute and each represent a probabilistic dependence. Efstathios, Charalambos, and Yannis (2007). Bayesian is based on hypothesis theory independency and it is hard to meet in reality. Jie and Hui(2008).

According to the Bayesian theorists, the hypothesis or prediction will be stronger of the sample size is smaller and the hypothesis will be weaker if the sample size is larger. Several test have been done to study Bayesian theory in detail. This particularly concerns the market securities since accounting and finance handles with large sets of data. Even though certain studies proves that Bayesian theory could produce accurate prediction but this accuracy only applicable to small sets of data. In conjunction with this issue it was proven that Bayesian theorem will not suite to the audit application since audit deals with huge database. D.J Johnstone (1990).

The need of data mining in audit

The widely spread news of the business disasters around the world has indirectly affect the public confidence in capital market system. The collapse of Enron and Arthur Andersen are one of the major impacts to business environment. Companies such as Siebel System, Qwest, WorldCom and Xerox are examples of companies which have problems with their bookkeeping. Royal Ahold which is known as Dutch retail Trade Company has also involved in this issue. Furthermore, UK Baring's Bank and the Japanese Daiwa's Bank have lost millions of dollars because of ineffective financial statement. All this issue of bankruptcy and financial fraud is because of inaccurate audit process.

Fact is that many parties, such as shareholders, investors, creditors, tax authorities, and managers are interested in the precision of organisations' financial performance. Auditors are in a position to monitor and control operations in organisations. Eija Koskivaara (2004).

The globally estimated average loss per organization from economic fraud is estimated to be $2,199,930 over a two-year period. According to Association of Certified Fraud Examiners (ACFE), approximately six percent of firm's revenues or $660 billion lost per year as the result of fraud. The continual increase in fraud has cause the rise up of several anti-fraud laws. However, many organizational anti-fraud efforts are not current and still have loopholes. Organizations are still finding ways to fight this fraud since red flags approach are not effective as well. Although red flag are associated with fraud it is still not perfect enough. Moreover, since it focuses attention on specific cues, it inhibits internal and external auditors from identifying other reasons of fraud occurrence. Another reason of organization finding different ways to fight back fraud is because entities use impractical way of fraud detection. It is also advisable for a firm to prevent fraud rather than take a lot of hassle to recover from it. Most of the companies and auditors deal with fraud on case-by-case basis rather than implementing it in long term plan. American Accounting Association (AAA) encourages researches to collaborate with auditors to overcome this issue by coming up with new technology or approach.

A research was conducted in order to identify how many organizations uses tools to combat fraud. Studies show that only 5.37 percent of organization uses data mining to fight fraud. James, Richard and Carl (2006).

Survey shows that only four percent of frauds were detected independently by auditors, the rest are discovered by accident. Fraud is hard to detect because it is a subject to falsification of accounting records. It is also related to the perpetrator's responses to the auditors' queries during the audit process. There are possibilities that he or she provide false or incomplete information when the auditor enquiry about the fraud related transaction. Furthermore, risk assessment will also affect the design of audit tests. When an auditor handles with high risk audit project they tend to perform more audit tests than necessary. This unnecessary test may reduce the efficiency of audit but the audit process would be the effective one. On the other hand, when auditors handles with low risk audit project they will tend to reduce the amount of audit test. This cause the overall effectiveness of the audit reduces. It will also lead to material misstatement and fraud may go undetected. Jerry, Mark and Jack (2005).

There are certain potential that the auditors may not discover material misstatements in the firm's financial statement. The likelihood of not discovering significant misstatement is called audit risk. During the audit process, internal control structure will be evaluated. The purpose of the internal control structure is to prevent or detect erroneous, fraudulent and missing accounting statements. A comprehensive and proper evaluation of this structure is significant to a successful audit. The effort to build intelligence based systems to shore up audit judgment tasks, such as control risk assessment (CRA), is reliant on the acquisition together with representation of knowledge and heuristics gained through audit experience. General audit theory and heuristics can be gained and represented using logical constructs. However, acquiring knowledge from expert auditors for decision-making between situations in the form of specific rules will be nearly hard and impossible. Assessing internal control structure somehow rather need hundreds of variables with thousands of possible inter relationship. It is not a step-by-step method but rather reacting based on experience by recognizing the patterns. Evaluating such complex relationships is a difficult task even for the most experienced auditors. That is why neural network systems is suggested to be used to automate judgment tasks that require this pattern recognition.

Fundamentally there are two audit approaches. In the substantive approach, the auditor will disregard the controls in place and focuses their analysis directly towards the information represented in the accounting statements. In opposition, in the control approach the auditor focuses first on the analysis of the internal control structure of the client in an effort to reduce the amount of substantive testing that will follow. Jefferson, Anne, Ronald (1997).

There are several flaws of financial scandals, which lead to audit failures.

The first flaw is the betrayal of CEOs' and CFOs' of neglecting the shareholders interest and primarily give importance to their own interest. Financial analyst, investments bankers, CPAs and other entities involved in this process. Executives who own large stock options benefit most from the increase in their company's stock price and it appears that these executives do some fraud in order to increase their company's stock price. This can be clearly seen when Enron's Jeffery Skilling made $112 million from stock option three years before Enron collapse. The second flaw is the apparent collusion among those involved in issuing securities and those involved in financial reporting. Citigroup was charged to pay $400 million because of issuing fraudulent research reports of allocating initial public share offerings to executives of other companies. The third flaw is the conflict of interest regarding for whom the company's auditor work. The USA enforced the S-O Act, in the event to clarify that auditors owe a primary loyalty to stockholders and to limit the conflict of interest that may result when two audit firms performs audit for a company. The fourth flaw is the loophole in Generally Accepted Accounting Principals (GAAP) itself. Some companies tend to use the special principals in GAAP for other wrong purpose. Dennis and Blair (2005).

Audit errors occurs when the organization fail to provide sufficient information to auditors. This is because organizations it self face a big issue in analyzing and act effectively on the information since they are dealing with a large amount of data. This will lead to difficulties in selecting the proper set of data in evaluating the financial statement of the company. Data accessing, including mining systems are gradually becoming essential to the organizations that wish to develop operational and other available data to improve the quality of decision making. Vijayan, Ranjit (1999).

Audit fraud occurs in four different areas. Those are material error, management fraud or misinterpretation, going concern and control risk assessment. Studies have been carried forward to incorporate ANN in those areas. Material error is one of the foremost areas where artificial neural network was applied widely. Material error applications direct auditors' interest towards financial account information where the actual relationships are not consistent with the expected relationships. Auditors have to decide what kind of further audit investigation they have to proceed with in order to explain the unexpected results.

Management fraud is known as an on purpose fraud committed by the management that affect the investors and creditors through misleading financial statements. As an auditor one cannot assume that the management is honest or dishonest. Auditors should take a hard, cold look at the possibility of management fraud or misinterpretation. This should be done at the start of the audit process and re-evaluate the likelihood of management fraud through out the audit process.

Going concern and financial distress requires the auditor to examine whether there is an extensive doubt about a client's ability to continue as a going concern for at least one year beyond the balance-sheet data. Going concern and financial distress is one of the most popular ANN research areas in business.

Auditors give opinion regarding going concern, when the company is at risk of failure or exhibits other signs of financial distress that question its ability to continue as a going concern. There are different types of financial distress and the auditors do have a choice between two types of going concern report. Those are modified audit report and disclaimer audit report. Bankruptcy is a situation where there is no ability to continue business and the auditor gives an unqualified audit report. The decision to issue a going concern opinion is an unstructured task that requires the use of the auditors' judgement.

Control risk assessment and audit fee. Auditors deals with massive amount of data when assessing the risk of the internal control structure. This will lead them failing to prevent or detect significant misstatements in financial statements. The relationships between internal control variables that must be identified, selected, and analysed often make examining a control risk a big hassle. Control risk assessment is a systematic process for integrating professional judgements about relevant risk factors. Their relative significance and undesirable conditions leads to identification of auditable activities. Eija Koskivaara (2004).

Data mining technique is very important in today's audit era. Auditing is used in determining companies' financial statement. It is also used to determine the cash flow of the company. The present study is conducted on application of data mining technique in audit which will help in the bankruptcy prediction. The economic crisis that occurred through East Asia in 1997 leads many companies to bankruptcy. The main cause of this issue is because of decision making problem that arose between auditors and accountants. Studies that were conducted to determine the relationship between external environment and firm's response towards them shows that economic crisis affect the way financial institutions operate. This indirectly affects the way firms evaluate their cash flows and financial statements so that they can minimize losses coming from bankruptcy. Furthermore under different environmental condition, decision makers in a firm are exposed to different types of decision making constraints. Since different people use different models and method for decision making, each of them come up with different conclusions. Using data mining techniques, the decision making is done based on the same rule and constraints given any kind on environmental issue such as crisis. In this way, the decision is more accurate, valid and reliable. Hian and Chan (2004)

Data mining as an auditing tool

The first thing that an auditor will look into when it comes to auditing is the companies' financial statement. Hence, the companies' management tend to do some fraud in order to modify the companies' financial statements. Fraudulent financial statements have become increasingly frequent for the past few years. One of the highlighting issues related to fraudulent financial statement was the Athens Stock Exchange in Greece where listed companies try to reduce their taxation on profits. So it is important for auditors to estimate the possibility of management fraud in an organization. The application of data mining technique for financial classification is an important research. Several law enforcement and special investigation units have implement data mining as an audit tool to identify fraudulent activities.

Financial statements are classified as audit fraud when there is an inclusion in the auditor's report of serious uncertainties as to the truthfulness of the accounts, taxation issue, incorrect balance sheet or income statement and any other misleading accounting issues. Sometimes high debt issues may also reflect to audit fraud or fraudulent financial statement because it shifts the risk from equity owners and managers to debt owners. This will lead will lead managers to manipulate the financial statement due to their need to meet debt covenants. Efstathios, Charalambos, and Yannis (2007).

One of the possible ways to overcome audit fraud and audit mistakes is by using a proper audit tools. Using data mining as an auditing tool can provide a good solution for this issue. Two approach is suggested as an data mining audit tool in this paper. Those are; (1) Neural network as an artificial intelligence, (2) combination of Artificial neural network (ANN) with analytical review (AR).

Neural Network as an artificial intelligence

Neural network is also known as one of the artificial intelligence technology. In recent years artificial intelligence is becoming a common technology with industrial application in accounting field. American Institute of Certified Public Accountant (AICPA) defines artificial intelligence as a technique which gives computer the human like abilities.

Neural networks are example based process and the knowledge in neural network is derived from data. Neural network have been proven as one of the important data mining technique in accounting and audit field. The application of neural network can be seen in real world situation when Avco Financial Service analyses loan application with neural network. There are also other examples such as neural networks created by Adaptive System assessing mortgage loan applications, Nestor and Wyman's neural network analyses on consumer loan applications and application of neural network by Chase Manhattan Bank to detect credit card fraud. Beside that, Nikko Securities, Frontier Financial and Citicorp improve their securities trading strategies by using neural network. This shows how far neural network have influenced the organizations. Furthermore, an audit firm was reported to be developing neural network system in order to identify audit errors. What is meant by audit errors is actually accounting miscalculation. A company should incorporate data mining technique such as neural network in order get accurate financial statement.

Many accounting task require expertise to give opinions concerning complex numeric and symbolic information. Moreover, many accounting task have added attraction of having high cost of wrong decisions. It is suggested that neural network is needed in external and internal auditing. Amelia A. Baldwin-Morgan (1995).

Neural network systems are applied in several studies to predict fraud litigation for assisting accountants on audit strategy making. Studies shows that the famous data mining technique, neural network provides promising predictive accuracy with better detecting power. It also tends to reduce misclassification of audit judgment and it is an artificial intelligence technique that is very well good in identifying fraud lawsuit. Unfortunately, financial scandal is an on going hazard in today's world. In the event of overcoming this issue, neural network approach have been applied in several audit fields such as assessment of material, evaluation of management fraud, issue of audit opinion, prediction financial crises, assessment of internal control system and decision of audit fee.

Neural network has the characteristic of stop training or learning when its performance is better than the benchmark in terms of performance criteria. Evidence from the research shows that neural network has an inexpensive cost in misclassification compared to logit. Studies also show that not only auditors face difficulty in audit decision-making but also neural network, which outperform CPA subject personal judgment.

The main purpose of the back propagation training in neural network is to obtain the weight of each node to minimize the squared error sum among the actual and the predicted value. First each edge will be given a random value. Then the squared error sum between the actual value and the predicted value will be calculated. Then the weight is recognized according to the gradient search method until the squared error sum is less than or equal to the predicted value. Finally obtained model is the one that has been trained and can then be used to forecast. Hsueh-Ju, Shaio-Yan and Chun-Long (2008).

Artificial Neural Network and Analytical Review

Artificial neural network (ANN) can be embedded in analytical review (AR) procedure for an accurate data mining audit tool. Analytical review procedures use comparison and relationships approach to assess whether account balances and other data are reasonable. Analytical review are performed predominantly at any of the three phases planning, testing, and completion during an audit process.

AR procedure works by performing comparisons on the recorded amounts and ratios with the expectations developed by the auditors. Auditors develops expectations by identifying and using possible relationships that are reasonably expected to exist based on their understanding of the industry in which the client operates. Analytical review procedures use several strategy and techniques to improve the efficiency of audits by developing these expectations and by comparing them with recorded amounts. The use of analytical review procedures entails the evaluation of the accuracy of account balances without considering the particulars of the individual transactions, which make up the account balance. AR plays an important role in assisting the auditor in determining the timing, nature and extent of the auditors substantive testing. It also helps in forming an overall opinion as to the reasonableness of recorded account values.

Artificial neural network (ANN) is a procedure that guide auditors in creating expected values and these expectations will then be compared with actual values automatically. ANN has several advantageous aspects if compared with other procedure. They are adaptive tools for processing data, which can learn, remember, and compare multifarious patterns. ANN can even recognise patterns in data, even when the data are noisy, ambiguous, distorted, or variable. It is argued that, ANN learns thru examples and then it will generalise the learning to new annotations.

Artificial neural networks are data driven and unlike traditional statistical techniques, ANN capable of identifying and simulating nonlinear relationships. Therefore, one significant advantage of artificial neural network could be that they provide additional information to the decision making process among auditors. ANN will help auditor to find hidden data during audit process. Moreover, ANN has been considered one of the up coming technologies. Information technology development and the high processing capacities of computers have made it possible to model ANN based information systems for monitoring and controlling operations. Eija Koskivaara (2004).

Artificial Neural Network (ANN), is an appropriate method used to detect management fraud in audit process. ANN has the great ability to identify nonlinear and independent relationship data. AutoNet are incorporated in ANN by using the generalized adaptive neural network (GANNA) framework. GANNA is similar like standard neural network. The advantage of GANNA is its speed and self determination of the correct network structure. The only thing which differs from ordinary neural network is that the network in AutoNet grows in complexity to solve the current problem. AutoNet has the ability of not using previous nodes or layers specification. This will increase the speed of data mining process. AutoNet uses simple quadratic function called squared error metric in an evolutionary manner. Several tests were conducted using the ANN approach in order to identify the effectiveness of audit results and fraud detection. ANN offer superior ability to standard methods in detecting FFS. The use of AutoNet in ANN offers advantage of quickly developing models for analysis. Alternative models such as Adaptive Logit Networks also offer additional potential for audit fraud.

The studies of ANN in data mining are very beneficial for the auditors. Auditors face a limited amount of time in handling audit process. They also need to study large data sets of financial and accounting statements in order to continue with their audit process. Applying data mining variables such as ANN and AutoNet methods will be able to lessen the auditors work and best able to detect FFS. Kurt and Kenneth (1998).

How data mining help auditors in audit process

Audit frauds are very hard to detect under normal audit procedure. This is because people tend to come up with different type of frauds as time passes by. Each time auditors may have to face new types of financial frauds until they could not find any errors when conducting the audit process. This is because they have limited knowledge concerning the characteristic of audit fraud. Moreover, most of the auditors' lack of experience to detect the fraud and sometimes the manager of the organization purposely tries to trick the auditors.

A research was conducted by Eining, Jones and Loebbecke in order to examine if the use of data mining method will improve the auditor's performance. They found that data mining techniques help auditors to deal with even more critical audit fraud and auditors was able to make a consistent decision regarding appropriate audit actions. Another research conducted by Green and Choi concerning neural network have been also evaluated. The results shows that neural network have significant capabilities when use it as fraud detection tool. Efstathios, Charalambos, and Yannis (2007).

Furthermore, the role of the auditors was a big question mark in audit field for the past few years. This is because as an auditor one should not only know how to evaluate the financial statement of an organization but also must consider other knowledge regarding audit fraud. They do not know which type of fraud is about to happen during the audit process. As an auditor one should exercise professional scepticism and professional care by questioning and critically assess the audit evidence. They should also develop their knowledge, skill and ability of specific assigned responsibility in order for them to commensurate with the identified risk. However, this is not an easy task to achieve since there are always different kinds of fraud from different kind of field. The question is how they can overcome this issue? Is the only way is to traditional method of identifying audit risk? Furthermore, the revolutionary of world is an on going process and there will be always an issue of fraud in audit.

There are also several audit standards which have been formed by Auditing Standards Board (ASB) in order to help and guide the auditors. Jayalakshmy, Seetharaman and Tan (2005).

On the other hand, auditors sometimes feel doubtful to take the decision of their hands and leave it to the data mining technology to make decision, but they should keep in mind that data mining technology is just to help auditors ease their difficulties in handling large data. They can use this tool as a comparison with their decision. It will also help them to gain better understanding of the characteristic, benefits and limitations of the financial data sets. Jefferson, Anne and Ronald (1997).


Limitation of artificial neural network is that the internal structure of ANN makes it difficult to trace the process of how the output was reached. This is why ANN lacks of clear explanation. Furthermore, internal process of the results is difficult to obtain because the connections' weights do not usually have obvious interpretations. However, from the auditing point of view, this black box problem is not an issue for the auditors.


Data mining technique has gained pervasive attention and increasing reputation in the business world. Successful data mining applications in modern years have been reported and recent surveys have found that data mining has growing usage and effectiveness.

Potentially data mining approaches such as artificial neural network will provide several benefits to audit firms. Data mining brings a new insights into the decision making process where it will guide auditors in decision aiding support and increase productivity.

Neural network system performs pretty well and might be a supportive tool for auditors. This implies that an ongoing innovation in artificial intelligence and data mining is needful as it could be employed to facilitate the evaluation of audit evidence.

The incorporation of data mining technology into decision support system will provide a more effective tool for auditors. Tools such s artificial neural network can help improve accuracy of audit decision by reducing bias or fault associated with the weighing and combining of fraud risk factors.