US 12,456,074 B2
Incremental machine learning for a parametric machine learning model
Si Er Han, Xian (CN); Ji Hui Yang, Beijing (CN); Xiao Ming Ma, Xian (CN); Jing Xu, Xian (CN); and Xue Ying Zhang, Xian (CN)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Nov. 4, 2021, as Appl. No. 17/453,540.
Prior Publication US 2023/0137184 A1, May 4, 2023
Int. Cl. G06N 20/00 (2019.01); G06F 18/20 (2023.01); G06F 18/21 (2023.01); G06F 18/232 (2023.01)
CPC G06N 20/00 (2019.01) [G06F 18/217 (2023.01); G06F 18/232 (2023.01); G06F 18/285 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
processing samples comprising historical samples and new samples with an existing parametric machine learning model to obtain at least one prediction residual of each of the samples, wherein the existing parametric machine learning model was trained based on the historical samples;
clustering the samples based on the at least one prediction residual of each of the samples and features of each of the samples;
sampling samples in each cluster to ensure that each cluster includes substantially similar number of sampled samples; and
updating the existing parametric machine learning model to obtain an updated parametric machine learning model based on sampled samples in each cluster, wherein the updating comprises:
training, for each specific cluster in all clusters, a temporary parametric machine learning model with the training sampled samples in all clusters except for those in the specific cluster to obtain a cluster-based parametric machine learning model;
evaluating, for each specific cluster in all clusters, the cluster-based parametric machine learning model with historical testing sampled samples and new testing sampled samples, respectively;
selecting from all cluster-based parametric machine learning models, cluster-based parametric machine learning models whose prediction accuracy for historical testing sampled samples is higher than a predefined threshold; and
selecting from the selected cluster-based parametric machine learning models, a specific cluster-based parametric machine learning model with the highest prediction accuracy for the new testing sampled samples within the selected cluster-based parametric machine learning models.