Automatic generation of optimized aggregation metrics for usage based insurance
Automatic generation of optimized aggregation metrics for usage-based insurance (UBI) includes performing a high-dimensional Bayesian optimization on a data archive, the data archive including a plurality of signals from vehicles and corresponding UBI effects, wherein the high-dimensional Bayesian optimization includes performing testing on weighted groups of the plurality of signals using one or more aggregation functions; and transmitting the one or more aggregation functions to a vehicle to cause the vehicle to provide aggregated signals for input to a UBI model to predict a UBI rate for the vehicle.
1 . A method for automatic generation of optimized aggregation metrics for usage-based insurance (UBI), comprising:
performing a multi-dimensional Bayesian optimization on a data archive, the data archive including a plurality of signals from vehicles and corresponding UBI effects, wherein the multi-dimensional Bayesian optimization includes performing testing on weighted groups of the plurality of signals using one or more aggregation functions, wherein the multi-dimensional Bayesian optimization is repeated until a convergence criterion is met and/or until an iteration budget is exhausted;
transmitting the one or more aggregation functions to a vehicle for execution by an on-vehicle processor of the vehicle, to cause the vehicle to provide aggregated signals generated from the plurality of signals without providing the plurality of signals, the aggregated signals being for input to a UBI model to predict a UBI rate for the vehicle;
receiving the aggregated signals from the vehicle, thereby reducing data transfer from the vehicle as compared to the vehicle providing the plurality of signals;
adding the aggregated signals to the data archive to update the data archive; and
repeating performing the multi-dimensional Bayesian optimization and the transmitting of the one or more aggregation functions to the vehicle based on the data archive as updated.
2 . The method of claim 1 , wherein the multi-dimensional Bayesian optimization includes:
training a surrogate model using input-output pairs to approximate system behavior;
formulating an acquisition function based on predictions of the surrogate model, the acquisition function being used to determine additional query points for the system behavior;
optimizing the acquisition function to obtain an input trigger;
evaluating the UBI model, using the obtained input trigger to obtain an actual UBI-relatedness output associated with the input trigger, thereby creating a new input-output pair; and
adding the new input-output pair to the data archive for further iterations.
3 . The method of claim 2 , wherein the acquisition function is based on mean and uncertainty from surrogate model predictions and is configured to maximize expected improvement.
4 . The method of claim 1 , wherein each of the one or more aggregation functions applies respective weights to each of the plurality of signals available from the vehicle, each respective weight being descriptive of a relative importance of a respective one of the plurality of signals to system behavior.
5 . The method of claim 1 , wherein weights below a threshold value are set to zero to excluded those of the plurality of signals, thereby reducing search space for the multi-dimensional Bayesian optimization and quantity of the aggregation functions.
6 . The method of claim 1 , further comprising:
receiving the aggregated signals from the vehicle based on the one or more aggregation functions; and
utilizing the aggregated signals to predict the UBI rate for the vehicle.
7 . A system for automatic generation of optimized aggregation metrics for UBI, comprising:
a data archive including a plurality of signals from vehicles and corresponding UBI effects; and
one or more computing devices configured to:
perform a multi-dimensional Bayesian optimization on the data archive, wherein the multi-dimensional Bayesian optimization includes performing testing on weighted groups of the plurality of signals using one or more aggregation functions, wherein the multi-dimensional Bayesian optimization is repeated until a convergence criterion is met and/or until an iteration budget is exhausted;
transmit the one or more aggregation functions to a vehicle for execution by an on-vehicle processor of the vehicle, to cause the vehicle to provide aggregated signals generated from the plurality of signals without providing the plurality of signals, the aggregated signals being for input to a UBI model to predict a UBI rate for the vehicle;
receive the aggregated signals from the vehicle, thereby reducing data transfer from the vehicle as compared to the vehicle providing the plurality of signals;
add the aggregated signals to the data archive to update the data archive; and
repeat performing the multi-dimensional Bayesian optimization and the transmit of the one or more aggregation functions to the vehicle based on the data archive as updated.
8 . The system of claim 7 , wherein the multi-dimensional Bayesian optimization includes to:
train a surrogate model using input-output pairs to approximate system behavior;
formulate an acquisition function based on predictions of the surrogate model, the acquisition function being used to determine additional query points for the system behavior;
optimize the acquisition function to obtain an input trigger;
evaluate the UBI model, using the obtained input trigger to obtain an actual UBI-relatedness output associated with the input trigger, thereby creating a new input-output pair; and
add the new input-output pair to the data archive for further iterations.
9 . The system of claim 8 , wherein the acquisition function is based on mean and uncertainty from surrogate model predictions and is configured to maximize expected improvement.
10 . The system of claim 7 , wherein each of the one or more aggregation functions applies respective weights to each of the plurality of signals available from the vehicle, each respective weight being descriptive of a relative importance of a respective one of the plurality of signals to system behavior.
11 . The system of claim 7 , wherein weights below a threshold value are set to zero to excluded those of the plurality of signals, thereby reducing search space for the multi-dimensional Bayesian optimization and quantity of the aggregation functions.
12 . The system of claim 7 , wherein the one or more computing devices are further configured to:
receive the aggregated signals from the vehicle based on the one or more aggregation functions; and
utilize the aggregated signals to predict the UBI rate for the vehicle.
13 . The system of claim 7 , wherein the one or more computing devices include a cloud server, and further comprising the vehicle, wherein the vehicle includes one or more controllers configured to:
aggregate the plurality of signals according to the one or more aggregation functions to generate the aggregated signals; and
send the aggregated signals to the cloud server without providing the plurality of signals to the cloud server.
14 . The system of claim 13 , wherein the one or more controllers of the vehicle are further configured to:
receive updated aggregation functions from the cloud server based on training of the UBI model by the cloud server; and
aggregate the plurality of signals according to the updated aggregation functions to generate additional aggregated signals; and
send the additional aggregated signals to the cloud server.
15 . A non-transitory computer-readable medium comprising instructions for automatic generation of optimized aggregation metrics for UBI that, when executed by one or more computing devices, cause the one or more computing devices to perform operations including to:
perform a multi-dimensional Bayesian optimization on a data archive, the data archive including a plurality of signals from vehicles and corresponding UBI effects, wherein the multi-dimensional Bayesian optimization includes performing testing on weighted groups of the plurality of signals using one or more aggregation functions, wherein the multi-dimensional Bayesian optimization is repeated until a convergence criterion is met and/or until an iteration budget is exhausted, wherein the multi-dimensional Bayesian optimization includes to:
train a surrogate model using input-output pairs to approximate system behavior,
formulate an acquisition function based on predictions of the surrogate model, the acquisition function being used to determine additional query points for the system behavior,
optimize the acquisition function to obtain an input trigger,
evaluate a UBI model, using the obtained input trigger to obtain an actual UBI-relatedness output associated with the input trigger, thereby creating a new input-output pair, and
add the new input-output pair to the data archive for further iterations;
transmit the one or more aggregation functions to a vehicle for execution by an on-vehicle processor of the vehicle, to cause the vehicle to provide aggregated signals generated from the plurality of signals without providing the plurality of signals, the aggregated signals being for input to the UBI model to predict a UBI rate for the vehicle;
receive the aggregated signals from the vehicle, thereby reducing data transfer from the vehicle as compared to the vehicle providing the plurality of signals;
add the aggregated signals to the data archive to update the data archive; and
repeat performing the multi-dimensional Bayesian optimization and the transmit of the one or more aggregation functions to the vehicle based on the data archive as updated.
16 . The non-transitory computer-readable medium of claim 15 , wherein the acquisition function is based on mean and uncertainty from surrogate model predictions and is configured to maximize expected improvement.
17 . The non-transitory computer-readable medium of claim 15 , wherein each of the one or more aggregation functions applies respective weights to each of the plurality of signals available from the vehicle, each respective weight being descriptive of a relative importance of a respective one of the plurality of signals to system behavior.
18 . The non-transitory computer-readable medium of claim 15 , wherein weights below a threshold value are set to zero to excluded those of the plurality of signals, thereby reducing search space for the multi-dimensional Bayesian optimization and quantity of the aggregation functions.
19 . The non-transitory computer-readable medium of claim 15 , further comprising instructions that, when executed by the one or more computing devices, cause the one or more computing devices to perform operations including to:
receive the aggregated signals from the vehicle based on the one or more aggregation functions; and
utilize the aggregated signals to predict the UBI rate for the vehicle.