US 12,455,173 B2
Training machine learning models based on movement and timing data
Matthew Louis Nowak, Midlothian, VA (US); Michael Anthony Young, Jr., Henrico, VA (US); and Christopher McDaniel, Glen Allen, VA (US)
Assigned to Capital One Services, LLC, McLean, VA (US)
Filed by Capital One Services, LLC, McLean, VA (US)
Filed on Nov. 15, 2022, as Appl. No. 18/055,665.
Prior Publication US 2024/0159555 A1, May 16, 2024
Int. Cl. G06N 20/00 (2019.01); G01C 21/00 (2006.01)
CPC G01C 21/383 (2020.08) [G01C 21/3856 (2020.08); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A system for map generation based on movement and timing data, the system comprising:
one or more processors; and
one or more non-transitory, computer-readable media comprising instructions that when executed by the one or more processors cause operations comprising:
receiving, from a first user device, an entry of training data for training a machine learning model to predict object locations within an indoor environment, wherein the entry of the training data comprises location data, timing data, and motion data collected based on a user input to a checklist application indicating whether a particular object has been checked off from a digital checklist at a time a user located the particular object, and wherein the location data indicates a device location of the first user device corresponding to the user input, the timing data indicates a timestamp corresponding to the user input, and the motion data indicates motion measurements of the first user device corresponding to the user input;
generating a precision parameter for the device location by assigning an initial value to the entry of the training data;
lowering the precision parameter for the device location based on the timing data and the motion data indicating that the user input was performed at some time other than the time the user located the particular object,
wherein the precision parameter is lowered by a first amount based on the timing data indicating that a timing interval between the user input and a previous or next user input being faster than a threshold time, and the precision parameter is lowered by a second amount based on the motion data indicating a rate of motion indicative of device velocity exceeding a threshold rate;
updating the training data with the precision parameter;
training, using the training data, the machine learning model to predict the object locations within the indoor environment;
receiving, from a second user device, a list comprising an object;
inputting the object into the machine learning model to generate an output indicating a location of the object within the indoor environment; and
outputting, to the second user device and using the output indicating the location of the object within the indoor environment, a map of the indoor environment showing a layout of the object.