US 12,394,300 B1
Providing location-based safety alerts
Michael Clinton Arndt, Valrico, FL (US); Rosa Maria Smith, San Antonio, TX (US); Nolan Serrao, Plano, TX (US); Olamide Oluwatomi Fanilola, Frisco, TX (US); Ana Rosa Maldonado, San Antonio, TX (US); and Megan Sarah Jennings, San Antonio, TX (US)
Assigned to UIPCO, LLC, San Antonio, TX (US)
Filed by UIPCO, LLC, San Antonio, TX (US)
Filed on Apr. 20, 2022, as Appl. No. 17/724,778.
Claims priority of provisional application 63/177,537, filed on Apr. 21, 2021.
This patent is subject to a terminal disclaimer.
Int. Cl. G08B 25/01 (2006.01); G08B 25/08 (2006.01); G08B 27/00 (2006.01); H04W 4/02 (2018.01)
CPC G08B 25/016 (2013.01) [G08B 25/08 (2013.01); G08B 27/005 (2013.01); H04W 4/025 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method, performed by a computing system, for alerting users to safety-related issues for geographic locations through a transformation of geographic location characteristic data into models and application of the models to geographic location specifics, the method comprising:
receiving geographic location characteristic data, wherein the geographic location characteristic data comprises, for corresponding geographic locations, danger conditions including one or more of: health statistics, safety statistics, weather statistics, or a combination thereof;
generating a danger score model, that includes one or more neural network machine learning models, that is trained to receive an indication of a geographic location and produce a danger score for the given geographic location, by building the danger score model using multiple training items, each training item corresponding to one of the instances of the geographic location characteristic data,
wherein each training item, corresponding to a particular geographic location, is created by:
computing a danger score for the geographic location using the health statistics, safety statistics, weather statistics, or a combination thereof, from the corresponding geographic location characteristic data; and
pairing the particular geographic location to the computed danger score; and
wherein building the danger score model includes training the one or more neural network machine learning models by, for each particular training item of the multiple training items:
providing input, to the one or more neural network machine learning models, based on the particular geographic location of the particular training item;
in response to providing the input to the one or more neural network machine learning models, receiving from the one or more neural network machine learning models a predicted danger score for the particular geographic location of the particular training item;
computing a loss function value, the loss function value indicating a difference between the predicted danger score and the danger score paired to the input, that was provided to the one or more neural network machine learning models and that was based on the particular geographic location of the particular training item; and
based on the loss function value, modifying aspects of the one or more neural network machine learning models;
identifying one or more geographic areas of interest associated with a user;
generating one or more danger scores by applying, to the danger score model, geographic location characteristic data, for at least the one or more geographic areas of interest, wherein the geographic location characteristic data, for at least the one or more geographic areas of interest, includes one or more of: health statistics, safety statistics, weather statistics, or a combination thereof; and
determining at least one of the one or more danger scores exceed one or more corresponding risk thresholds, and in response:
generating a safety alert based on the determination; and
transmitting the safety alert to the user.