Ivan Callo – Revenue Assurance & Data Science Engineer

  • „Modern Revenue Assurance System must be robust to process millions of records
    but agile for fast configuration of reconciliation rules.”
    “Even that is not enough, Unsupervised Machine Learning must be designed
    as First line of defense.”
                                                                                                                      Ivan Callo




    Revenue Leakage Detection System

    Revenue Assurance System loads data from several source Databases, reconciliates loaded data to detect customer/provisioning/billing discrepancies and provides discrepancies to Data Analysts for further investigation and raising of Trouble Tickets. Key Performance Indicators are presented below.

    Revenue Assurance System including data loading, data reconciliation and Trouble Ticket Reporting.

    Machine Learning in Revenue Assurance

    Unsupervised Machine Learning doesn´t need particular knowledge about processed data due to unsupervised learning feature. That means, there is no need for preparing of training and testing data sets. (This is a fundamental difference from rule-based reconciliation where each rule must be defined precisely for specific type of alarm.) Therefore Machine Learning is outstanding in simplicity of implementation and speed of deployment. A perfect tool for First line of Defense.

    There is an example of unsupervised Machine Learning implementation based on Local Outlier Factor. This method is looking for exceptions/spikes/outliers in processed records. The standard records are identified by the value 1, but each deviation from the standard is defined by the value -1. The X score defines the distance from the standard, the larger the absolute value of the X score, the greater the deviation from the standard.

    Unsupervised Machine Learning in Revenue Assurance System

    © 2026 Ivan Callo