Systems and methods for determining a risk score using machine learning based at least in part upon collected sensor data
A computer-implemented method for determining risk scores using machine learning based on sensor data. The method includes receiving sensor data and demographic data for a user, determining a pattern of behavior based on the sensor data and demographic data, identifying a plurality of test groups based on other users that exhibit similar patterns of behavior to each other where sensor data from other users' electronic devices are received, identifying a test group corresponding to the user from among the plurality of test groups, determining a risk score using a trained risk analyzer machine learning model trained with historical sensor data, historical demographic data, and feedback data indicative of incidents, identifying risk mitigation strategies based on the risk score, refining the risk mitigation strategies using an iterative machine learning process as additional feedback data is collected, and retrieving risk mitigation communications from a database when the risk score exceeds a threshold.
1 . A computer-implemented method for operating a computing device comprising:
receiving, by one or more processors and from one or more sensors, sensor data corresponding to a user;
receiving, by the one or more processors, demographic data corresponding to the user;
determining a pattern of behavior of the user based on the sensor data and the demographic data;
identifying a plurality of test groups based on other users that exhibit similar patterns of behavior to each other, wherein other sensor data from electronic devices of the other users are received to identify the plurality of test groups;
identifying a test group corresponding to the user from among the plurality of test groups based on the determined pattern of behavior of the user;
determining, using a trained risk analyzer machine learning model, a risk score of the user based at least in part upon the identified test group corresponding to the user, wherein the trained risk analyzer machine learning model is trained using training data comprising historical sensor data and historical demographic data from a plurality of users and feedback data indicative of incidents corresponding to risk;
identifying one or more risk mitigation strategies based at least on the determined risk score of the user;
refining, using an iterative machine learning process, the one or more risk mitigation strategies as additional feedback data indicative of incidents is collected; and
retrieving, from a database, one or more risk mitigation communications to display to the user when the determined risk score of the user exceeds a threshold.
2 . The computer-implemented method of claim 1 further comprising:
determining the one or more risk mitigation strategies to implement for the user based in part on the determined risk score of the user.
3 . The computer-implemented method of claim 2 , wherein the one or more risk mitigation strategies further comprise at least one of: a guidance communication, a recommendation to perform an action, or a reminder to use a mitigation strategy displayed after a time period.
4 . The computer-implemented method of claim 1 further comprising:
after retrieving one of the one or more risk mitigation communications, transmitting the one of the one or more risk mitigation communications to a user interface of an electronic device of the user, wherein the one of the one or more risk mitigation communications comprises a text message, a numerical risk score, a percentage, an illustration, or an icon.
5 . 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, by the one or more processors and from one or more sensors, sensor data corresponding to a user;
receiving, by the one or more processors, demographic data corresponding to the user;
determining a pattern of behavior of the user based on the sensor data and the demographic data;
identifying a plurality of test groups based on other users that exhibit similar patterns of behavior to each other, wherein other sensor data from electronic devices of the other users are received to identify the plurality of test groups;
identifying a test group corresponding to the user from among the plurality of test groups based on the determined pattern of behavior of the user;
determining, using a trained risk analyzer machine learning model, a risk score of the user based at least in part upon the identified test group corresponding to the user, wherein the trained risk analyzer machine learning model is trained using training data comprising historical sensor data and historical demographic data from a plurality of users and feedback data indicative of incidents corresponding to risk;
identifying one or more risk mitigation strategies based at least on the determined risk score of the user;
refining, using an iterative machine learning process, the one or more risk mitigation strategies as additional feedback data indicative of incidents is collected; and
retrieving, from a database, one or more risk mitigation communications to display to the user when the determined risk score of the user exceeds a threshold.
6 . The system of claim 5 , wherein the operations further comprise:
determining the one or more risk mitigation strategies to implement for the user based in part on the determined risk score of the user.
7 . The system of claim 6 , wherein the one or more risk mitigation strategies further comprise at least one of: a guidance communication, a recommendation to perform an action, or a reminder to use a mitigation strategy displayed after a time period.
8 . The system of claim 5 , wherein the operations further comprise:
after retrieving one of the one or more risk mitigation communications, transmitting the one of the one or more risk mitigation communications to a user interface of an electronic device of the user, wherein the one of the one or more risk mitigation communications comprises a text message, a numerical risk score, a percentage, an illustration, or an icon.
9 . One or more non-transitory computer-readable media storing computing instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
receiving, by the one or more processors and from one or more sensors, sensor data corresponding to a user;
receiving, by the one or more processors, demographic data corresponding to the user;
determining a pattern of behavior of the user based on the sensor data and the demographic data;
identifying a plurality of test groups based on other users that exhibit similar patterns of behavior to each other, wherein other sensor data from electronic devices of the other users are received to identify the plurality of test groups;
identifying a test group corresponding to the user from among the plurality of test groups based on the determined pattern of behavior of the user;
determining, using a trained risk analyzer machine learning model, a risk score of the user based at least in part upon the identified test group corresponding to the user, wherein the trained risk analyzer machine learning model is trained using training data comprising historical sensor data and historical demographic data from a plurality of users and feedback data indicative of incidents corresponding to risk;
identifying one or more risk mitigation strategies based at least on the determined risk score of the user;
refining, using an iterative machine learning process, the one or more risk mitigation strategies as additional feedback data indicative of incidents is collected; and
retrieving, from a database, one or more risk mitigation communications to display to the user when the determined risk score of the user exceeds a threshold.
10 . The one or more non-transitory computer-readable media of claim 9 , wherein the operations further comprise:
determining the one or more risk mitigation strategies to implement for the user based in part on the determined risk score of the user.
11 . The one or more non-transitory computer-readable media of claim 10 , wherein the one or more risk mitigation strategies further comprise at least one of: a guidance communication, a recommendation to perform an action, or a reminder to use a mitigation strategy displayed after a time period.
12 . The one or more non-transitory computer-readable media of claim 10 , wherein determining the risk mitigation strategies comprises selecting communications from a risk mitigation database, and wherein the communications comprise at least one of: videos, navigation directions, links to applications, or phone numbers for services.
13 . The one or more non-transitory computer-readable media of claim 9 , wherein the operations further comprise:
after retrieving one of the one or more risk mitigation communications, transmitting the one of the one or more risk mitigation communications to a user interface of an electronic device of the user, wherein the one of the one or more risk mitigation communications comprises a text message, a numerical risk score, a percentage, an illustration, or an icon.
14 . The one or more non-transitory computer-readable media of claim 9 , wherein the sensor data comprises at least one of: global positioning system (GPS) data, accelerometer data, gyroscope data, heart rate data, or compass data.
15 . The one or more non-transitory computer-readable media of claim 9 , wherein the risk score comprises at least one of: a risk of illness, a risk of accident, a risk of injury, a risk of death, a risk of mental health issues, a risk of exposure, or a composite risk score.
16 . The one or more non-transitory computer-readable media of claim 9 , wherein the operations further comprise:
periodically receiving updated sensor data corresponding to the user; and
determining an updated risk score of the user based at least in part upon the updated sensor data.
17 . The one or more non-transitory computer-readable media of claim 9 , wherein the pattern of behavior comprises at least one of: time spent at one or more locations, frequency of visits to the one or more locations, time of day when at the one or more locations, driving routes, or driving speeds.
18 . The one or more non-transitory computer-readable media of claim 9 , wherein the one or more sensors are housed in at least one of: a smartphone, a tablet computer, a wearable computing device, or a device embedded in a vehicle.
19 . The one or more non-transitory computer-readable media of claim 9 , wherein the feedback data indicative of incidents corresponding to risk comprises data relating to at least one of: accidents, deaths, insurance claims, illnesses, or injuries from users in the plurality of test groups.
20 . The one or more non-transitory computer-readable media of claim 9 , wherein the operations further comprise:
receiving additional feedback data indicative of incidents corresponding to risk after determining the risk score;
generating, using an adjustment function, adjustment operations based on a metric assessing a difference between actual and expected outputs of the trained risk analyzer machine learning model; and
applying the adjustment operations to the trained risk analyzer machine learning model to refine future risk score determinations.