US 11,935,075 B2
Card inactivity modeling
Akash Singh, Delhi (IN); Tanmoy Bhowmik, Bangalore (IN); Deepak Bhatt, Dehradun (IN); Shiv Markam, Mandla (IN); Ganesh Nagendra Prasad, West New York, NJ (US); and Jessica Peretta, Purchase, NY (US)
Filed by MASTERCARD INTERNATIONAL INCORPORATED, Purchase, NY (US)
Filed on Aug. 10, 2021, as Appl. No. 17/398,387.
Claims priority of application No. 202011034927 (IN), filed on Aug. 13, 2020.
Prior Publication US 2022/0051269 A1, Feb. 17, 2022
Int. Cl. G06Q 30/0201 (2023.01); G06F 18/2415 (2023.01); G06N 20/00 (2019.01)
CPC G06Q 30/0201 (2013.01) [G06F 18/2415 (2023.01); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A system of account inactivity modeling through hierarchical data models including a first level classifier and a second level classifier, comprising:
a processor programmed to:
access feature data comprising a plurality of features associated with each account from among a plurality of accounts;
access first label data that represents an occurrence of attrition;
train, based on the feature data and the first label data, the first level classifier to generate, for each account, a first output indicating whether or not the account is likely to exhibit attrition;
access second label data comprising a plurality of labels comprising at least a first label indicating attrition will start at a first time period and a second label indicating that attrition will start at a second time period different than the first time period;
train, based on the feature data and the plurality of labels, the second level classifier to generate, for each account determined by the first level classifier to exhibit attrition, a prediction of a start of the attrition;
generate one or more evaluation metrics that indicate performance of the first level classifier and/or the second level classifier trained on the feature data;
identify a subset of features from the feature data based on the one or more evaluation metrics to improve model performance and/or reduce the amount of data processed by the first level classifier and/or the second level classifier; and
re-train the first level classifier and/or the second level classifier based on the identified subset of features.