US 12,333,397 B1
Systems and methods for incorporating supplemental shape information in a support vector machine
Geng Deng, McLean, VA (US); Yaoguo Xie, Charlotte, NC (US); Xindong Wang, Dayton, MD (US); and Qiang Fu, McLean, VA (US)
Assigned to Wells Fargo Bank, N.A., San Francisco, CA (US)
Filed by Wells Fargo Bank, N.A., San Francisco, CA (US)
Filed on Feb. 1, 2021, as Appl. No. 17/164,405.
Int. Cl. G06N 20/10 (2019.01); G06F 18/211 (2023.01); G06F 18/2411 (2023.01); G06F 18/2431 (2023.01); G06F 18/25 (2023.01)
CPC G06N 20/10 (2019.01) [G06F 18/211 (2023.01); G06F 18/2411 (2023.01); G06F 18/2431 (2023.01); G06F 18/25 (2023.01)] 18 Claims
OG exemplary drawing
 
1. A method for data classification using a shape-restricted support vector machine (SR-SVM), the method comprising:
receiving a training dataset;
selecting, by a training engine, a set of shape restrictions, the set of shape restrictions including a shape restriction for each feature in the training dataset, wherein selecting the set of shape restrictions comprises:
identifying, by the training engine, a first set of shape restrictions, the first set of shape restrictions including a shape restriction for each feature in the training dataset, and
for a particular feature in the training dataset,
generating an approximation spline function based on the shape restriction for the particular feature in the first set of shape restrictions; and
in an instance in which the approximation spline function has a slope of zero, replacing the shape restriction for the particular feature in the first set of shape restrictions with a different shape restriction for the particular feature;
training, by the training engine, the SR-SVM using the training dataset and the selected set of shape restrictions, wherein training the SR-SVM produces a shape-restricted hyperplane that defines a decision boundary separating a first class of data points in the training dataset from a second class of data points in the training dataset;
receiving a target data point;
using the trained SR-SVM to classify the target data point into a first classification or a second classification; and
outputting an indication of whether the target data point is in the first classification or the second classification.