Anomalies are transactions which are suspicious because they differ significantly from standard behaviors or patterns. We'll be exploring various methods of anomaly detection and their application in audit field work, utilizing the following:
- Benford's Law
- Exploratory Data Analysis
- Unsupervised Machine Learning
- Principal Component Analysis
- Clustering
- Outlier Detection
The dataset consists of 1k executed purchase orders transactions, this is a fictitious data and it was actually fun discovering some suspicious patterns without forcing any. Our objective is to perform guided sampling, that is to make a selection of purchase orders according to specific criteria, instead of doing random sample selection.