US 12,169,785 B2
Cognitive recommendation of computing environment attributes
Indervir Singh Banipal, Austin, TX (US); Shikhar Kwatra, San Jose, CA (US); Nadiya Kochura, Bolton, MA (US); and Sourav Mazumder, Contra Costa, CA (US)
Assigned to INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Apr. 15, 2021, as Appl. No. 17/231,639.
Prior Publication US 2022/0335302 A1, Oct. 20, 2022
Int. Cl. G06N 3/088 (2023.01); G06N 3/044 (2023.01); G06N 3/045 (2023.01); G06N 5/01 (2023.01)
CPC G06N 3/088 (2013.01) [G06N 3/044 (2023.01); G06N 3/045 (2023.01); G06N 5/01 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
generating, using a training dataset, a decision tree, wherein each node of the decision tree represents a question sequence, wherein each question sequence comprises a plurality of questions intended to elicit a response usable in recommending a first configuration of a plurality of computing resources, wherein the training dataset comprises a plurality of asset characteristics of assets within previously configured computing environments;
receiving a user response to a first question sequence, wherein the first question sequence is represented by a first node of the decision tree;
generating, by inputting the user response to a recursive neural network (RNN), a second question sequence and a first deviation value, the first deviation value indicative of a deviation between the second question sequence and a plurality of decision tree question sequences, each decision tree question sequence in the plurality of decision tree question sequences represented by a next node in the decision tree, each next node in the decision tree connected by an edge to the first node;
disambiguating, using a disambiguation rule, the user response, the disambiguating performed responsive to determining that the first deviation value exceeds a threshold, the disambiguating resulting in a disambiguated user response;
generating, by inputting the user response to the RNN, a third question sequence and a second deviation value, the second deviation value indicative of a deviation between the third question sequence and the plurality of decision tree question sequences;
generating, using a generator portion of a generative associative network (GAN), a custom question sequence;
generating, using a user response to the custom question sequence, a projected usage parameter; and
generating, using the projected usage parameter, the first configuration of the plurality of computing resources.