US 12,292,297 B2
Computational model for creating personalized routes based at least in part upon predicted total cost of claim frequency or severity
Kenneth Jason Sanchez, San Francisco, CA (US)
Assigned to Quanata, LLC, San Francisco, CA (US)
Filed by QUANATA, LLC, San Francisco, CA (US)
Filed on Jul. 7, 2023, as Appl. No. 18/219,430.
Application 18/219,430 is a continuation of application No. 16/883,436, filed on May 26, 2020, granted, now 11,852,495.
Prior Publication US 2023/0349706 A1, Nov. 2, 2023
Int. Cl. G01C 21/34 (2006.01); G01C 21/36 (2006.01); G06N 3/044 (2023.01); G06N 3/08 (2023.01); G06Q 40/08 (2012.01); G06Q 50/26 (2012.01)
CPC G01C 21/3484 (2013.01) [G01C 21/3461 (2013.01); G01C 21/3691 (2013.01); G06N 3/044 (2023.01); G06N 3/08 (2013.01); G06Q 40/08 (2013.01); G06Q 50/265 (2013.01)] 21 Claims
OG exemplary drawing
 
1. A computer-implemented method, performed by one or more processors, for determining one or more personalized risk values associated with a driving route, the computer-implemented method comprising:
determining, by a trained machine learning model comprising a recurrent neural network, base risks associated with a set of road segments, wherein the recurrent neural network is trained using at least labeled training data, wherein the labeled training data comprises a frequency or a severity of incidents reported associated with operation of vehicles corresponding to the set of road segments, and wherein during training, outputs from at least one layer of the recurrent neural network are fed back to at least one previous layer to identify inputs for determining the base risks or intermediate layer outputs, wherein the recurrent neural network is further trained to determine one or more weights for at least one input comprising a road type, a road condition, and a driving behavior, and wherein the one or more weights are modified based at least in part upon a personalized driver profile;
receiving one or more driving routes corresponding to a driver, wherein each driving route of the one or more driving routes comprises the set of road segments; and
determining, using the recurrent neural network, as trained, for each driving route of the one or more driving routes, a respective personalized risk value based on analyzing (i) the personalized driver profile and (ii) the base risks associated with the set of road segments.