Data Mining and Accounting - Case Study Example

Overview

Data mining, also known as knowledge discovery in databases KDD or data, involves data analysis technique in which data is analyzed from different angles to come up with recognized patterns that can be interpreted into useful information that can be applied to solve problems or to help in operations. It simply involves searching for useful information in large data banks and detecting patterns that exist between different data sets and applying those predictions in learning new information that can be applied to a situation (Kumar & Bhardwaj, 2011). Data mining is an important tool that has been applied for many years in accounting. In accounting, data mining uses different analytical tools and software to analyze data efficiently and generate any relationship that may exist for interpretation. It draws large amounts of data in the data warehouse and defines their correlations that are interpreted into useful information. With the advancement in information technology and computer knowledge, data mining has become a very necessary tool in any accounting field.

Birth of Data Mining in Accounting

Data mining as a technique is not new in the accounting field rather it has been in practice ever since before even the advent of computers where important information was being made from a handful of clumsy data ranging from large sheets of entries and log books. During this period, frauds were hard to detect due to clumsy nature of business transactions. The early life of data mining began by simple techniques of statistics and machine learning. Data analysis applications during this period were very tiresome, slow and expensive. Data mining came into being in late 1980 where some concepts from the subjects of pattern recognition, intelligence, traditional statistical probability as well as database knowledge began to construct the new term ‘data mining’ It received its name from the ‘mining’ useful information in large data warehouses and also became known as knowledge discovery which has been its main (Koyuncugil & Ozgulbas, 2011).  In 1982 Pawlak introduced a theory known as Rough Set, which was one of the methods used in data mining (Kumar & Bhardwaj, 2011).

With advanced innovation, in the late 1980’s in computer technology, many powerful computers were developed that had excellent processing capabilities, enormous storage abilities as well as complex software to run various programs. Companies and large organizations were already in possession of large data banks and data warehouses. These companies were in dire need of using these powerful computers in sifting important information from these huge data banks. Some of the remarkable examples of this include simple techniques such as scanning supermarket data volumes as well as analysis of various market reports over a given period.

As innovations continued, complex terminologies like knowledge discovery in databases (KDD) took shape in the 1990s, and it was data processing term that referred more or less data mining (Kumar & Bhardwaj, 2011).During this period, analysts were working hard towards developing important methods known as algorithms which could enable companies to search, analyze and handle voluminous amounts of data. Algorithms technique, therefore, became an important pioneering tool in data mining. There was a big leap in data mining when IBM Company developed Weka database miner which was also called intelligent miner in 1995. This innovation was useful because it is technically simplified algorithms and supported KDD analysis of any fraudulent activities that was going on in various corporations at that time (Kumar & Bhardwaj, 2011).

The model that really created the modern day data mining technique was established in the year 2000. This year is seen as the breakthrough year by all data mining technologists because it was when the first standard called Cross Industry Standard Process (CRISP) was developed and it was an effective model that had set of tasks, outputs, and inputs. The system was also able to come up with metrics, and this made it a desired process model that could handle all Company’s large volumes of data (Tan, Steinbach, & Kumar, 2014). Therefore statistical and machine learning techniques are fingerprint techniques from this era employing numerical data obtained in large data banks kept by Companies to areas of interest to acquire useful information.

Data Mining and Fraud Today

The world of information technology and computer science developed rapidly giving rise to vast fields of knowledge, and this made distinguish between data mining and KDD. According to Koyuncugil & Ozgulbas (2011), KDD is all the processes of discovering useful information contained in a given volume of data, while application of algorithms in coming up with detected patterns from data. In this definition, they posit that data mining is a combination of past and present achievements in statistics, machine learning as well as database techniques. Currently, data mining technique is highly developed and customized to be used on different platforms such as mainframes, servers as well as personal computers. They also come in different prices which are cheaper compared to the prices in the early 1990s. Artificial intelligence AI, statistical methods, machine learning and pattern re-arrangement was advancing now, and the data that are generated are heterogeneous meaning that they can be structured, semi-structured or unstructured data. This shows that not only is there a change in types of data that can be mined, but also the type of networks have also evolved to include resources such as end storage, high-speed networks as well as high-tech computing technologies.

Management Fraud identification

Data mining technique is an important tool today in the identification of fraud committed by the management which causes a lot of loss to the organization and inflicts financial harm to all stakeholders including government tax authorities. Two methods famous in management fraud identification include z-score and logistic regression which can identify accurately any manipulation done on financial statements by the managers which are separated from financial distress by the technique (Kumar & Bhardwaj, 2011).

Fraud comprises of numerous financial actions ranging from bribery, management theft, hacking, bankruptcy, theft of identity of business fraud. Accountants know how to distinguish between internal and external fraud. For example, when someone steals from a company he is said to have committed internal fraud while engaging in stealing for a company is known as external fraud. Data mining can be used by fraud detectors or auditors to discover analyze and report fraud to the responsible authorities.

Most of the common data mining techniques used to detect fraud include digital analysis, outlier detection, and trending.

Digital analysis

This is a method of data mining for the fraud that does not regard context of the data. It comes up with patterns if the routine is violated and thus makes the auditor to question the transaction. This method uses Benford’s law of predicting unusual numbers and activities. This has been a famous method with most auditors in the recent years (Williams, 2011). 

Outlier detection

Applies a z-score application of data mining technique which involves dividing value minus mean with standard deviation (Williams, 2011). This application is helpful to auditors especially in identifying fraud committed by credit card. Also supervised and unsupervised machine learning can identify the same fraud by looking at spending rate to detect abnormal spending as well as frequency of spending on that particular credit card.

Trending

Data mining can compare different numbers over a period and detect when the trend increases exponentially. The computer programs will show which items and values are increasing through linear regression as well as the goodness of fit slope (Rahman, 2009). Other software like Statistical Program for Social Sciences SPSS and packages like SAS are a useful tool in indicating trending values to auditors investigating fraud.

Auditing

This technique is used by many auditors to discover accounting malpractices which could otherwise be impossible to discover. Malpractices such as the famous ‘book cooking’ is a thing of the past considering the full application of data mining technique (Williams, 2011). Use of Analytical review techniques and other analytical procedures gives the modern auditor accurate account balance that is straight forward, and therefore data mining continues to break grounds in making voluminous data analysis possible. Other important data mining techniques used in auditing include; Neural Networks, Case Base Reasoning, and Genetic Algorithm. These applications of data mining technique are useful in conducting new risk auditing (Williams, 2011). 

Predicting Bankruptcy

Data mining is popularly applied in predicting bankruptcy, therefore, informing investors, creditors, employees as well as business managers on the financial status of those they are dealing with saving them from damages and unnecessary risks. Corporate failures can also be predicted precisely using Logistic regression and machine learning applications of data mining technique (Williams, 2011).  Through application of Neural Networks and Fuzzy Systems, the accountant can model a bankruptcy problem with uttermost precision.

Going into the Future

There should be more research in this field of data mining for fraud detection because as the development in technology increases so does the fraud techniques. The auditors also should be taught thorough computer applications dealing with data mining so as they can apply it in their daily activities. For instance use of Neural Networks, Fuzzy logic and genetic programming in large Corporation data warehouse can detect fraud instantly as well as project future fraudulent activities and should be embraced (Venkatadri & Lokanatha, 2011).

Conclusion

In the past, techniques employed included statistical and machine learning techniques which acted on numerical data and data that were found in traditional data banks. This past data mining technique found their application in business management fields such as accounting. This contrast sharply with the current trend where data mining employs AI and pattern reassembling in addition to the past techniques on various data which may include unstructured, structured and semi-structured formats. These techniques operate in networks with high speed, end storage, and advanced computing resources. The management of fraud in future will be an easy task for auditors because the software developed will effectively address the fraud before even its occurrence.