Local Outlier Factor (LOF) Tool
The Local Outlier Factor (LOF) tool provides an efficient and user-friendly solution for identifying anomalies in the transaction data of a specific agency. By applying the Local Outlier Factor (LOF) algorithm, the tool calculates an LOF score for each transaction, which indicates how significantly a transaction deviates from the normal patterns observed in the data. Transactions with higher LOF scores are flagged as potential outliers, suggesting unusual activity such as exceptionally high amounts or infrequent vendors.
The tool analyzes transaction data from the past 3 years, allowing users to focus on recent trends and patterns. It offers the flexibility to filter data based on specific criteria, such as time periods, vendors, or transaction amounts, enabling a more targeted analysis. Once the LOF scores are calculated, users can define a threshold to identify the most significant outliers. These results can then be exported in JSON or CSV formats, making it easy to integrate the findings into other tools or systems for further investigation.
The Local Outlier Factor (LOF) algorithm is particularly effective at detecting local outliers within dense clusters of data, making it well-suited for identifying subtle anomalies that might otherwise go unnoticed. By combining advanced analytics with customizable data extraction, the LOF tool empowers users to efficiently uncover and analyze suspicious transactions, enhancing their ability to monitor and address potential irregularities.
Typical Thresholds for Anomalies (LOF):
Strong Outliers: LOF Score ≥ 2.5
Moderate Outliers: LOF Score between 1.5 and 2.5
Normal Data Points: LOF Score < 1.5