US 12,093,873 B2
User performance analysis and correction for S/W
Sun Young Park, San Diego, CA (US); Srivenkata Laksh Gantikota, San Diego, CA (US); Dustin Michael Sargent, San Diego, CA (US); and Marwan Sati, Mississauga (CA)
Assigned to International Business Machines Corporation, Armonk, NY (US)
Filed by INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed on Jan. 22, 2021, as Appl. No. 17/156,123.
Prior Publication US 2022/0237540 A1, Jul. 28, 2022
Int. Cl. G06Q 10/0639 (2023.01); G06N 20/00 (2019.01); G06Q 10/0631 (2023.01); G16H 30/20 (2018.01)
CPC G06Q 10/0639 (2013.01) [G06N 20/00 (2019.01); G06Q 10/063112 (2013.01); G16H 30/20 (2018.01)] 19 Claims
OG exemplary drawing
 
1. A system for optimizing user interaction with a software application, the system comprising:
an electronic processor configured to:
receive a collection of interaction data including a plurality of interaction sequences and data associated with each of the plurality of interaction sequences, wherein each of the interaction sequences relates to user interactions with a computer user interface when performing a task for a software function;
determine, in a training operation of an artificial intelligence (AI) model, a user skill level performance metric representing the user's level of expertise for each of the plurality of interaction sequences according to an identified ground truth, wherein the performance metric is a value assigned to each of the plurality of interaction sequences representing a skill level of users performing each of the plurality of interaction sequences;
train, with the collection of interaction data and the skill level performance metric determined for each of the plurality of interaction sequences as training data for performing supervised learning, the AI model to output a performance rating for a received interaction sequence, wherein the AI model is a time-series deep leaning model selected from the group consisting of a recurrent neural network model, a long short-term memory model, and a time-based convolution network model, and the training is provided to the time-series deep learning model, and wherein AI model is trained to learn an optimal sequence of interactions performed by an expert user by mapping each interaction sequence in the collection of interaction data to time and accuracy for each interaction sequence;
receive a current interaction sequence of a user for the software application, apply the trained AI model to the current interaction pattern of the user to assign a performance rating to the current interaction sequence of the user based on comparing the current interaction sequences of the user to the optimal sequence of interactions learnt by the AI model during training;
generate, via the AI model as applied to the current interaction pattern of the user, a recommendation for display within a user interface, for a modified user interaction pattern of the user for the software application based on the assigned performance rating;
output feedback along with the recommendation to the use upon analyzing the current interaction sequence of the user;
and retrain the AI model using the feedback.