SYSTEMS AND METHODS FOR IMPROVING DRIVER SAFETY USING UPLIFT MODELING
A computer-implemented method can include receiving a first set of telematics data collected by one or more sensors associated with a vehicle of a user. The computer-implemented method can also include determining a driving tip to be provided to the user based at least on the first set of telematics data. The computer-implemented method can further include generating a treatment for providing the driving tip to the user. The computer-implemented method can additionally include providing the driving tip to the user in accordance with the treatment. The computer-implemented method can also include receiving a second set of telematics data collected by the one or more sensors after providing the driving tip. The computer-implemented method can further include determining whether the user complied with the driving tip based at least on the second set of telematics data. The computer-implemented method can additionally include updating the machine learning model. Other embodiments are described.
1 . A computer-implemented method comprising:
receiving a first set of telematics data collected by one or more sensors associated with a vehicle of a user;
determining a driving tip to be provided to the user based at least on the first set of telematics data;
generating a treatment for providing the driving tip to the user, the treatment being generated using a machine learning model and comprising a plurality of treatment facets, wherein generating the treatment comprises selecting an option from a plurality of options for each treatment facet of the plurality of treatment facets by applying an uplift modeling technique and by combining (i) a respective weight associated with the option based at least on an effectiveness of the option for users within a set of users, and (ii) a respective confidence score associated with the option based at least on a sample size of treatments containing the option provided to the users within the set of users;
providing the driving tip to the user in accordance with the treatment;
receiving a second set of telematics data collected by the one or more sensors after providing the driving tip;
determining whether the user complied with the driving tip based at least on the second set of telematics data; and
updating the machine learning model by at least adjusting the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets.
2 . The computer-implemented method of claim 1 , wherein at least one of:
(a) the machine learning model:
comprises a plurality of possible channels of delivery, timings of delivery, and tip contents; and
associates each of the plurality of possible channels of delivery, timings of delivery, and tip contents with a respective initial score;
(b) when the user is determined to have complied with the driving tip, adjusting the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets comprises:
increasing the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets utilized in the treatment; or
(c) when the user is determined to have not complied with the driving tip, adjusting the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets comprises:
decreasing the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets utilized in the treatment.
3 . The computer-implemented method of claim 1 , further comprising:
determining a respective initial score for each of a plurality of possible channels of delivery, timings of delivery, and tip contents based at least on demographic data of the user and an effectiveness of each of the plurality of possible channels of delivery, timings of delivery, and tip contents for other users having similar demographic data as the user.
4 . The computer-implemented method of claim 3 , wherein the other users having similar demographic data are located within a geographic range of the user.
5 . The computer-implemented method of claim 2 , wherein tip contents of the plurality of possible channels of delivery, timings of delivery, and tip contents comprise a monetary amount, a likelihood of damage or injury, a comparison to other users, or an indication of performance over time.
6 . The computer-implemented method of claim 2 , wherein:
the machine learning model associates each of the plurality of possible channels of delivery, timings of delivery, and tip contents with a corresponding confidence level based at least on a corresponding sample size associated with a number of treatments having at least one of the plurality of possible channels of delivery, timings of delivery, or tip contents provided to other users having similar demographic data as the user.
7 . The computer-implemented method of claim 6 , wherein generating the treatment comprises:
assigning respective weights to the plurality of possible channels of delivery, timings of delivery, and tip contents based at least on respective initial scores and respective confidence levels for the plurality of possible channels of delivery, timings of delivery, and tip contents; and
selecting a channel of delivery, a timing of delivery, and a tip content of the plurality of possible channels of delivery, timings of delivery, and tip contents based at least on the respective weights and the respective confidence levels, as assigned.
8 . A system comprising:
one or more processors; and
one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform operations comprising:
receiving a first set of telematics data collected by one or more sensors associated with a vehicle of a user;
determining a driving tip to be provided to the user based at least on the first set of telematics data;
generating a treatment for providing the driving tip to the user, the treatment being generated using a machine learning model and comprising a plurality of treatment facets, wherein generating the treatment comprises selecting an option from a plurality of options for each treatment facet of the plurality of treatment facets by applying an uplift modeling technique and by combining (i) a respective weight associated with the option based at least on an effectiveness of the option for users within a set of users, and (ii) a respective confidence score associated with the option based at least on a sample size of treatments containing the option provided to the users within the set of users;
providing the driving tip to the user in accordance with the treatment;
receiving a second set of telematics data collected by the one or more sensors after providing the driving tip;
determining whether the user complied with the driving tip based at least on the second set of telematics data; and
updating the machine learning model by at least adjusting the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets.
9 . The system of claim 8 , wherein at least one of:
(a) the machine learning model:
comprises a plurality of possible channels of delivery, timings of delivery, and tip contents; and
associates each of the plurality of possible channels of delivery, timings of delivery, and tip contents with a respective initial score;
(b) when the user is determined to have complied with the driving tip, adjusting the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets comprises:
increasing the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets utilized in the treatment; or
(c) when the user is determined to have not complied with the driving tip, adjusting the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets comprises:
decreasing the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets utilized in the treatment.
10 . The system of claim 8 , wherein the operations further comprise:
determining a respective initial score for each of a plurality of possible channels of delivery, timings of delivery, and tip contents based at least on demographic data of the user and an effectiveness of each of the plurality of possible channels of delivery, timings of delivery, and tip contents for other users having similar demographic data as the user.
11 . The system of claim 10 , wherein the other users having similar demographic data are located within a geographic range of the user.
12 . The system of claim 9 , wherein tip contents of the plurality of possible channels of delivery, timings of delivery, and tip contents comprise a monetary amount, a likelihood of damage or injury, a comparison to other users, or an indication of performance over time.
13 . The system of claim 9 , wherein:
the machine learning model associates each of the plurality of possible channels of delivery, timings of delivery, and tip contents with a corresponding confidence level based at least on a corresponding sample size associated with a number of treatments having at least one of the plurality of possible channels of delivery, timings of delivery, or tip contents provided to other users having similar demographic data as the user.
14 . The system of claim 13 , wherein generating the treatment comprises:
assigning respective weights to the plurality of possible channels of delivery, timings of delivery, and tip contents based at least on respective initial scores and respective confidence levels for the plurality of possible channels of delivery, timings of delivery, and tip contents; and
selecting a channel of delivery, a timing of delivery, and a tip content of the plurality of possible channels of delivery, timings of delivery, and tip contents based at least on the respective weights and the respective confidence levels, as assigned.
15 . A non-transitory computer-readable media storing computing instructions that, when executed on one or more processors, cause the one or more processors to perform operations comprising:
receiving a first set of telematics data collected by one or more sensors associated with a vehicle of a user;
determining a driving tip to be provided to the user based at least on the first set of telematics data;
generating a treatment for providing the driving tip to the user, the treatment being generated using a machine learning model and comprising a plurality of treatment facets, wherein generating the treatment comprises selecting an option from a plurality of options for each treatment facet of the plurality of treatment facets by applying an uplift modeling technique and by combining (i) a respective weight associated with the option based at least on an effectiveness of the option for users within a set of users, and (ii) a respective confidence score associated with the option based at least on a sample size of treatments containing the option provided to the users within the set of users;
providing the driving tip to the user in accordance with the treatment;
receiving a second set of telematics data collected by the one or more sensors after providing the driving tip;
determining whether the user complied with the driving tip based at least on the second set of telematics data; and
updating the machine learning model by at least adjusting the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets.
16 . The non-transitory computer-readable media of claim 15 , wherein at least one of:
(a) the machine learning model:
comprises a plurality of possible channels of delivery, timings of delivery, and tip contents; and
associates each of the plurality of possible channels of delivery, timings of delivery, and tip contents with a respective initial score;
(b) when the user is determined to have complied with the driving tip, adjusting the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets comprises:
increasing the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets utilized in the treatment; or
(c) when the user is determined to have not complied with the driving tip, adjusting the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets comprises:
decreasing the respective weight and the respective confidence score associated with the option, as selected, for each treatment facet of the plurality of treatment facets utilized in the treatment.
17 . The non-transitory computer-readable media of claim 15 , wherein the operations further comprise:
determining a respective initial score for each of a plurality of possible channels of delivery, timings of delivery, and tip contents based at least on demographic data of the user and an effectiveness of each of the plurality of possible channels of delivery, timings of delivery, and tip contents for other users having similar demographic data as the user.
18 . The non-transitory computer-readable media of claim 17 , wherein the other users having similar demographic data are located within a geographic range of the user.
19 . The non-transitory computer-readable media of claim 16 , wherein at least one of:
tip contents of the plurality of possible channels of delivery, timings of delivery, and tip contents comprise a monetary amount, a likelihood of damage or injury, a comparison to other users, or an indication of performance over time; or
the machine learning model associates each of the plurality of possible channels of delivery, timings of delivery, and tip contents with a corresponding confidence level based at least on a corresponding sample size associated with a number of treatments having at least one of the plurality of possible channels of delivery, timings of delivery, or tip contents provided to other users having similar demographic data as the user.
20 . The non-transitory computer-readable media of claim 19 , wherein generating the treatment comprises:
assigning respective weights to the plurality of possible channels of delivery, timings of delivery, and tip contents based at least on respective initial scores and respective confidence levels for the plurality of possible channels of delivery, timings of delivery, and tip contents; and
selecting a channel of delivery, a timing of delivery, and a tip content of the plurality of possible channels of delivery, timings of delivery, and tip contents based at least on the respective weights and the respective confidence levels, as assigned.