Chapter 14

Identifying Fraud Using Time-Series Analysis

A time-series analysis extrapolates the past into the future and compares current results to those predictions. Large deviations from the predictions signal a change in conditions, which might include fraud. A time-series is an ordered sequence of the successive values of an expense or revenue stream in units or dollars over equally spaced time intervals. Time-series is well suited to forensic analytics because accounting transactions usually include a time or date stamp. The main objectives with time-series analysis are to (a) give the investigator a better understanding of the revenues or expenditures under investigation, and, (b) to predict the revenues or expenses for future periods. These predicted values will be compared to the expected results and large differences investigated.

The comparison of actual to predicted results is closely related to a continuous monitoring setting. Time-series analysis has been made easier to use over the past few years by user-friendly software and the increased computing power of personal computers. An issue with time-series is that the diagnostic statistics are complex and this might make some forensic investigators uncomfortable in drawing conclusions that have forensic implications. For example, there are usually three measures to measure the accuracy of the fitted model and users might not know which measure is the best.

The usual forensic analytics application is forecasting the ...

Get Forensic Analytics: Methods and Techniques for Forensic Accounting Investigations now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.