Investigating accounting fraud patterns using data mining techniques

Abstract EN

A data mining technology was utilized to discover fraud trends in accounting dataset that contained fraudulent transactions.

The following method was used to achieve the goal: first, inside data, fraudulent transactions were settled according to three fraud patterns; then, using the Rapidminer program, the algorithms, Euclidian distance, and local outlier factors were run.

As a result, the deception was exposed.

Patterns were shown in a variety of ways depending on the visuals provided by the application.

Finally, using the k Means technique allowed for an effective group clustering of the data by Euclidian distance.

As a result of the distribution of values, the first and third frauds were discovered.

The outlier detection algorithm (LOF) correctly identified the three fraud behaviors caused by isolated outliers in diverse situations.