US 12,333,585 B2
Anomaly detection for bill generation
Rahul Yadav, Alwar (IN); Nidhi Rajput, Daman and Diu (IN); Veresh Jain, Bangalore (IN); Krishnapriya Jaya Radhakrishnan, Thiruvananthapuram (IN); Amit Omprakash Rathi, Pune (IN); Saroj Jasmine Merwana, Pune (IN); and Aashima Sethi, New York, NY (US)
Assigned to ORACLE INTERNATIONAL CORPORATION, Redwood Shores, CA (US); and ORACLE FINANCIAL SERVICES SOFTWARE LIMITED, Goregaon (IN)
Filed by Oracle International Corporation, Redwood Shores, CA (US); and Oracle Financial Services Software Limited, Goregaon (IN)
Filed on Mar. 31, 2022, as Appl. No. 17/710,745.
Claims priority of provisional application 63/293,024, filed on Dec. 22, 2021.
Prior Publication US 2023/0196420 A1, Jun. 22, 2023
Int. Cl. G06Q 30/04 (2012.01); G06Q 10/10 (2023.01)
CPC G06Q 30/04 (2013.01) [G06Q 10/10 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A method comprising:
receiving, by a bill processing system implemented using one or more computing systems, a request to verify a bill record associated with an entity account, the bill record stored in a memory of the bill processing system;
identifying, by the bill processing system, a bill level machine learning (ML) model to be used for verifying the bill record, the bill level ML model trained using a training algorithm and historical data comprising historical bills and associated bill level features;
detecting, by the bill processing system, based at least in part on the bill level ML model, bill level anomaly information for the bill record based on a spatial distribution of the historical data learned by the bill level model using the training algorithm and the bill record, wherein the bill level anomaly information comprises information about an anomaly detected in the bill record;
identifying, by the bill processing system, a bill line level ML model to be used for verifying one or more bill lines in the bill record, the bill line level ML model trained using the training algorithm, the historical data comprising the historical bills and associated bill line level features, wherein the training algorithm includes a supervised learning training algorithm or an unsupervised learning training algorithm;
detecting, by the bill processing system, based at least in part on the bill line level ML model, bill line level anomaly information for the bill record based on the spatial distribution of the historical data learned by the bill line level model using the training algorithm and the bill record, wherein the bill line level anomaly information comprises information about an anomaly detected at a bill line in the bill record;
receiving, by the bill processing system, a vector representation for verifying the bill record, the vector representation identifying a set of products that are expected to be included in a billing cycle of the bill record;
identifying, by the bill processing system, additional anomaly information associated with the bill record based on the vector representation;
aggregating, by the bill processing system, the bill level anomaly information detected by the bill level ML model, the bill line level anomaly information detected by the bill line level ML model, and the additional anomaly information to generate a bill verification report for the bill record;
providing, by the bill processing system, the bill verification report as a response to the request received to verify the bill record, the bill verification report identifying the additional anomaly information associated with the bill record, wherein the additional anomaly information identifies a product that is not in the set of products identified in the vector representation;
analyzing, by a feedback system that is communicatively coupled to the bill processing system, the bill verification report for the bill record; and
providing, by the feedback system, a result of the analysis, in real-time, to a training system used by the bill processing system to train a set of ML models for verifying a set of bill records associated with a set of entity accounts, wherein the set of ML models include at least the bill level ML model to be used for verifying the set of bill records and the bill line level ML model to be used for verifying the set of bill records.