US 11,816,542 B2
Finding root cause for low key performance indicators
Lukasz G. Cmielowski, Cracow (PL); Rafal Bigaj, Kracow (PL); Wojciech Sobala, Cracow (PL); and Maksymilian Erazmus, Zasow (PL)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Jan. 3, 2020, as Appl. No. 16/733,552.
Claims priority of application No. 19198001 (EP), filed on Sep. 18, 2019.
Prior Publication US 2021/0081833 A1, Mar. 18, 2021
Int. Cl. G06N 20/00 (2019.01); G06F 17/15 (2006.01); G06F 17/16 (2006.01); G06F 18/23213 (2023.01)
CPC G06N 20/00 (2019.01) [G06F 17/15 (2013.01); G06F 17/16 (2013.01); G06F 18/23213 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method for identifying a change of an indicator value for a system of interdependent entities, the method comprising:
determining, based on initial input data, an initial key performance indicator (KPI) value, the KPI value representing operational quality of a system of interdependent entities, the system of interdependent entities including a machine-learning model as a component of the interdependent entities;
logging additional input data for each entity during subsequent operation of the system of interdependent entities;
deriving, from the additional input data, scoring payload data and related results of the machine-learning model, the related results being classifiers of a neural network of the machine-learning model;
clustering, based on similarity, the additional input data into a plurality of defined clusters;
determining a set of cluster-based KPI values for each cluster of the plurality of defined clusters;
generating a vector for the set of cluster-based KPI values;
calculating metric values of the machine-learning model by mapping each cluster onto the scoring payload data, resulting in a metric values vector; and
determining correlation matrix values for a correlation matrix between the vector of cluster-based KPI values and the metric values vector of the clusters.