US 12,462,299 B2
Multi-layer artificial intelligence models for progressive predictions during data pipeline management
Haipeng Gao, San Jose, CA (US)
Assigned to PAYPAL, INC., San Jose, CA (US)
Filed by PayPal, Inc., San Jose, CA (US)
Filed on Mar. 15, 2024, as Appl. No. 18/607,394.
Application 18/607,394 is a continuation of application No. 17/537,267, filed on Nov. 29, 2021, granted, now 11,961,137.
Prior Publication US 2024/0303730 A1, Sep. 12, 2024
Int. Cl. G06Q 40/03 (2023.01); G06N 5/04 (2023.01); G06N 20/00 (2019.01)
CPC G06Q 40/03 (2023.01) [G06N 5/04 (2013.01); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A method of using a multi-layer predictive machine learning (ML) engine, comprising;
receiving, at a feature extraction module of the multi-layer predictive ML engine, a first set of data associated with an entity and usable as an input to the multi-layer predictive ML engine for time-based predictions at different times over a time period, and wherein the first set of data is associated with a first time during the time period;
segmenting, by the feature extraction module, the first set of data in accordance with a first set of ML features and a second set of ML features, wherein the first set of ML features remain unchanged over the time period and the second set of ML features dynamically change over the time period;
generating, by a first classifier of the multi-layer predictive ML engine trained on static feature data and connected to the feature extraction module, based on the segmented first set of data for the first set of ML features, a first ML model output indicating a static feature score associated with the first time;
generating, by a second classifier of the multi-layer predictive ML engine trained on dynamic feature data and connected to both the first classifier and the feature extraction module, based on the segmented first set of data for the second set of ML features, the first ML model output, a first time-based prediction indicating a first final feature score associated with the first time;
receiving a second set of data for the predictive ML engine associated with a second time after the first time during the time period;
segmenting, by the feature extraction module, the second set of data in accordance with the first set of ML features and the second set of ML features;
using the first ML model output as an input to the second classifier of the multi-layer predictive ML engine; and
generating, by the second classifier of the multi-layer predictive ML engine, based on the segmented second set of data for the second set of ML features, the first ML model output, a second time-based prediction indicating a second final feature score for the second time.