US 11,860,721 B2
Utilizing automatic labelling, prioritizing, and root cause analysis machine learning models and dependency graphs to determine recommendations for software products
Ravindra Kabbinale, Bangalore (IN); Sherin Varghese, Bangalore (IN); Santhosh MV, Kasaragod (IN); Bhavana V Gudi, Bengaluru (IN); Sneha S. Shekar, Bangalore (IN); Shruthi Dhivakaran, Bengaluru (IN); Rajendra Prasad Tanniru, Basking Ridge, NJ (US); Aditi Kulkarni, Bangalore (IN); Vijeth Srinivas Hegde, Bangalore (IN); and Koushik M. Vijayaraghavan, Chennai (IN)
Assigned to Accenture Global Solutions Limited, Dublin (IE)
Filed by Accenture Global Solutions Limited, Dublin (IE)
Filed on Jul. 20, 2021, as Appl. No. 17/380,763.
Prior Publication US 2023/0021373 A1, Jan. 26, 2023
Int. Cl. G06F 11/07 (2006.01); G06N 20/00 (2019.01); G06F 11/36 (2006.01); G06F 18/214 (2023.01); G06F 11/34 (2006.01)
CPC G06F 11/079 (2013.01) [G06F 11/0793 (2013.01); G06F 11/3466 (2013.01); G06F 11/3604 (2013.01); G06F 18/214 (2023.01); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A method, comprising: receiving, by a device, historical software data identifying events and logs associated with software products utilized by an entity; processing, by the device, the historical software data, with a data labelling model, to generate historical health scores, historical sentiment scores, and historical dissimilarity scores for the software products; combining, by the device, the historical health scores, the historical sentiment scores, and the historical dissimilarity scores to determine historical error severity scores for the software products; automatically training, by the device, a machine learning model, with the historical software data and the historical error severity scores, to generate a trained machine learning model, wherein the training comprises: determining, based on supervised learning of the machine learning model and based on new input, a prediction; implementing a feedback loop to train the machine learning model; and determining whether the prediction satisfies a threshold level of accuracy, and wherein new historical software data is processed with the trained machine learning model instead of the data labeling model when the prediction satisfies the threshold level of accuracy; receiving, by the device, software data identifying current logs and events associated with software products utilized by the entity; processing, by the device, the software data, with the trained machine learning model, to generate error severity scores for the software products; processing, by the device, the error severity scores, with a prioritization model, to generate prioritized error scores; processing, by the device, the error severity scores and the prioritized error scores, with a root cause analysis model, to generate root cause data identifying root causes associated with the error severity scores; performing, by the device, one or more actions based on the root cause data; receiving, by the device, feedback via the feedback loop; and retraining, by the device and based on the feedback, the trained machine learning model.