US 12,411,751 B2
Hierarchical modeling approach for digital repair parts prediction
Rômulo Teixeira de Abreu Pinho, Rio de Janeiro (BR); Adriana Bechara Prado, Rio de Janeiro (BR); Roberto Nery Stelling Neto, Rio de Janeiro (BR); Jeffrey Scott Vah, Round Rock, TX (US); Aaron Sanchez, Austin, TX (US); and Ravi Shukla, Bangalore (IN)
Assigned to Dell Products L.P., Round Rock, TX (US)
Filed by Dell Products L.P., Round Rock, TX (US)
Filed on Mar. 15, 2022, as Appl. No. 17/695,526.
Prior Publication US 2023/0315607 A1, Oct. 5, 2023
Int. Cl. G06F 11/34 (2006.01); G06F 18/243 (2023.01); G06N 3/0442 (2023.01); G06N 20/00 (2019.01); G06N 20/20 (2019.01)
CPC G06F 11/3476 (2013.01) [G06F 11/3428 (2013.01); G06F 18/24323 (2023.01); G06N 20/20 (2019.01); G06N 3/0442 (2023.01); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A method, comprising:
accessing input data comprising data elements from logs that identify user problems experienced with computing system components, the data elements each being associated with a respective original class label that identifies a class of computing system components to which the data element relates, the respective original class labels forming a group of class labels, and a first one of the original class labels is overrepresented in the group;
reducing the overrepresentation of the first original class label in the group by creating an arbitrary aggregation of some of the class labels that includes the first original class label;
building a hierarchical classification modelling structure configured to classify the input data using the aggregation, and also using one of the original class labels;
creating, based on a configuration of the hierarchical modeling structure, prepared data in which one or more of the original class labels is replaced by the aggregation;
training, using the prepared data, a hierarchical model that is included in the hierarchical classification modeling structure;
training a benchmark model using the original class labels;
collecting classification performance metrics of the benchmark model and of the hierarchical model;
generating a prediction, using the hierarchical model, to obtain a first predicted label;
generating a prediction, using the benchmark model, to obtain a second predicted label; and
comparing, based on the first predicted label and the second predicted label, the classification performance metrics of the benchmark model with the classification performance metrics of the hierarchical model.