Sunil Joshua Mathew

US Patent:

20210312458, Oct 7, 2021

Filed:

Jun 15, 2021

Appl. No.:

17/347940

Inventors:

- Mumbai, IN
Sunil J. MATHEW - Brooklyn NY, US

International Classification:

G06Q 20/40
G06N 20/00
G06K 9/62
G06Q 40/02

Abstract:

Systems, methods, and other embodiments for ML-Based automated below-the-line threshold tuning include, in one embodiment, training an ML model to predict probabilities that an event is fraudulent on a set of events (i) sampled from a set of historic events labeled by an alerting engine as either above-the-line events or below-the-line events on either side of a threshold line indicating that an event is suspicious, and (ii) confirmed to be either fraudulent or not fraudulent; determining that the alerting engine should be tuned based on differences between probability values predicted for the events by the trained machine learning model and the labels applied to the events; generating a tuned threshold value for the threshold line based at least in part on the probability values predicted by the machine learning model; and tuning the alerting engine by replacing a threshold value with the tuned threshold value to adjust the threshold line.

US Patent:

20200104849, Apr 2, 2020

Filed:

Sep 28, 2018

Appl. No.:

16/145952

Inventors:

- Mumbai, IN
Sunil J. MATHEW - Brooklyn NY, US

International Classification:

G06Q 20/40
G06Q 40/02
G06K 9/62
G06F 15/18

Abstract:

Systems, methods, and other embodiments associated with applying machine learning to below-the-line threshold tuning are described. In one embodiment, a method includes selecting a set of sampled events and labeling each event in the set of sampled events as either suspicious or not suspicious. Then, a machine learning model to calculate for a given event a probability that the given event is suspicious is built based on the set of sampled events. The machine learning model is trained, and its calibration validated. Based on probabilities calculated by the machine learning model, a scenario and segment combination to be tuned is determined. A tuned threshold value is generated, and an alerting engine is adjusted with the tuned parameter to reduce errors by the alerting engine in classifying events as not suspicious.