US 12,236,485 B2
Method and system for providing portfolio deviation analytics
Chandra Gorantla, Sugar Land, TX (US); and Sumit Khanna, East Windsor, NJ (US)
Assigned to JPMORGAN CHASE BANK, N.A., New York, NY (US)
Filed by JPMorgan Chase Bank, N.A., New York, NY (US)
Filed on Dec. 6, 2021, as Appl. No. 17/457,678.
Prior Publication US 2023/0177609 A1, Jun. 8, 2023
Int. Cl. G06Q 40/06 (2012.01); G06Q 40/04 (2012.01)
CPC G06Q 40/06 (2013.01) [G06Q 40/04 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A method for providing portfolio deviation analytics, the method being implemented by at least one processor, the method comprising:
generating, by the at least one processor, a predictive model by using an artificial neural network;
training, by the at least one processor using training data, the predictive model based on a model assessment method that includes at least one from among a holdout method, a K-fold-cross-validation method, and a bootstrap method;
assessing, by the at least one processor, the predictive model to determine whether a least squares error rate is within a predetermined range;
automatically deploying, by the at least one processor, the predictive model when the least squares error rate is within the predetermined range;
retrieving, by the at least one processor, at least one data set from at least one source based on a predetermined schedule for a geographic location that is delineated based on a shared characteristic, the at least one data set including at least one structured data set and at least one unstructured data set;
retrieving, by the at least one processor via an application programming interface, reference data from at least one reference data hub;
updating, by the at least one processor, the reference data by,
parsing, by the at least one processor, the reference data to determine whether a reference data element in the reference data is stale,
wherein the reference data is associated with a predetermined time threshold, and
wherein the reference data element is determined to be stale when an age of the reference data element exceeds the associated predetermined time threshold;
identifying, by the at least one processor, a data element in the at least one data set that corresponds to the reference data element; and
updating, by the at least one processor, the reference data element with the corresponding data element when the reference data element is stale;
parsing, by the at least one processor, the updated reference data to identify at least one portfolio and at least one corresponding parameter;
identifying, by the at least one processor, a deviation amount for each of the at least one portfolio based on the corresponding at least one data element and the at least one corresponding parameter;
determining, by the at least one processor using the automatically deployed predictive model, at least one resolution action for each of the at least one portfolio based on the corresponding deviation amount, the automatically deployed predictive model including a machine learning model that uses a historical pattern of a plurality of resolution actions; and
automatically initiating, by the at least one processor, the at least one resolution action.