Detecting Transaction Anomalies

While detecting transaction anomalies, the historical transaction data of the given period needs to be grouped based on certain parameters, such as division, transaction source, transaction record type. Only then, the system can detect more reliable anomalies at different levels. To enable the transaction anomaly detection feature in ORMB, you need to do the following:

  • Set the Transaction Anomaly Detection Feature Opt-In option type of the C1-AIMLFC feature configuration to Y.

  • Enable the transaction anomaly detection feature for the required divisions through the master configuration created using the C1-AIMLMST business object.

  • Set the various parameters, such as number of days for which historical data should be collected, column set for which anomalies can be detected at different levels, machine learning parameters, etc. through the C1-TADFC feature configuration.

A new batch named C1-TADPR is introduced in this release. It reads the historical transaction data from the CI_​TXN_​DETAIL table for the divisions for which the transaction anomaly detection feature is enabled. The system considers the transaction data for the period given in the C1-TADFC feature configuration. For example, if the number of days is set to 30 and batch business date is 01-Nov-2025, then system will consider the transaction data from 02-Oct-2025 till 31-Oct-2025. Once the historical data is extracted, the system groups the data based on the comma-separated columns given in the Grouping Columns parameter. It then identifies the unusual patterns in the transactions present in the CI_​TXN_​DETAIL_​STG table based on the historical grouped data. Based on the patterns, it detects anomalies (if any) in the columns given in the Monitored Columns parameter. These anomalies are stored in the C1_​TAD_​PRED and C1_​TAD_​DAY_​PRED tables. Once the batch is successfully executed, the following two reports are generated in the Excel format:

  • Transaction Detail Full Report - Lists whether there are any anomalies at the transaction level.

  • Transaction Day Anomaly Details Report - Lists whether there are any anomalies in the aggregated amount or count for the day.

Depending on whether the anomaly exists or not, the IS_​ANOMALY column is set to Y or N, respectively. If an anomaly exists for the transaction or day, the system indicates the reason for anomaly in the Reason Codes column. Here, the reason code is a concatenated string consisting of the column name and anomaly reason. The anomaly reason could VHIGH, HIGH, LOW, VLOW. However, if the insufficient data exists while detecting transaction anomalies, the IS_​ANOMALY column is set to X.