Collision analysis platform using machine learning to reduce generation of false collision outputs
Aspects of the disclosure relate to computing platforms that utilize machine learning to reduce false positive/negative collision output generation. A computing platform may apply machine learning algorithms on received data to generate a collision output. In response to generating the collision output indicating a collision, the computing platform may identify a data collection location. If the data collection location is within a predetermined radius of a false positive collection location, the computing platform may modify the collision output to indicate a non-collision. If the data collection location is not within the predetermined radius, the computing platform may compute a score using telematics data and compare the score to a predetermined threshold. If the score does not exceed the predetermined threshold, the computing platform may modify the collision output to indicate a non-collision. If the score exceeds the predetermined threshold, the computing platform may affirm the collision output indicating a collision.
1 . A system comprising:
at least one processor;
a communication interface communicatively coupled to the at least one processor; and
memory storing computer-readable instructions that, when executed by the at least one processor, cause the system to:
after receiving sensor data indicative of an occurrence of a collision and an indication that a data collection location corresponding to the sensor data is within a predetermined radius of a false positive collection location, determine that the collision did not occur;
after receiving an indication that the data collection location corresponding to the sensor data is not within the predetermined radius of the false positive collection location, determine a first likelihood of collision score;
after determining that the first likelihood of collision score does not exceed a predetermined collision threshold, determine that the collision did not occur; and
after determining that the first likelihood of collision score exceeds the predetermined collision threshold, determine that the collision did occur.
2 . The system of claim 1 , wherein the instructions are executable by the at least one processor to cause the system to:
analyze angular velocity data included in the sensor data to compute the first likelihood of collision score.
3 . The system of claim 2 , wherein the instructions for causing the system to analyze the angular velocity data are further executable by the at least one processor to cause the system to: compare the angular velocity data to one or more machine learning datasets corresponding to non-collision events in which a mobile computing device was dropped.
4 . The system of claim 1 , wherein the instructions for causing the system to receive the sensor data indicative of the occurrence of the collision are further executable by the at least one processor to cause the system to:
receive sensor data from one or more vehicle based sensors.
5 . The system of claim 1 , wherein the instructions for causing the system to receive the sensor data indicative of the occurrence of the collision are further executable by the at least one processor to cause the system to:
receive sensor data from one or more mobile computing devices; and
receive location data indicative of respective locations of the one or more mobile computing devices.
6 . The system of claim 5 , wherein the instructions are executable by the at least one processor to cause the system to:
communicate a corroboration request to a first mobile computing device of the one or more mobile computing devices;
after receiving an indication from a user of the first mobile computing device that the collision occurred, communicate an indication of the collision to an emergency dispatch system.
7 . The system of claim 1 , wherein the false positive collection location comprises one or more of: a ski resort, an amusement park, or a body of water.
8 . The system of claim 1 , wherein the instructions for causing the system to receive the sensor data indicative of the occurrence of the collision are further executable by the at least one processor to cause the system to:
apply one or more machine learning algorithms to the sensor data, wherein the one or more machine learning algorithms are trained to infer, based on the sensor data, whether a collision occurred;
wherein the instructions are further executable by the at least one processor to cause the system to:
after determining that the first likelihood of collision score does not exceed a predetermined collision threshold, update the one or more machine learning algorithms to indicate that a false positive collision determination was made by the one or more machine learning algorithms.
9 . A non-transitory computer-readable medium having stored thereon computer-readable instructions that, when executed by at least one processor of a system, cause the system to:
after receiving sensor data indicative of an occurrence of a collision and an indication that a data collection location corresponding to the sensor data is within a predetermined radius of a false positive collection location, determine that the collision did not occur;
after receiving an indication that the data collection location corresponding to the sensor data is not within the predetermined radius of the false positive collection location,
determine a first likelihood of collision score;
after determining that the first likelihood of collision score does not exceed a predetermined collision threshold, determine that the collision did not occur; and
after determining that the first likelihood of collision score exceeds the predetermined collision threshold, determine that the collision did occur.
10 . The non-transitory computer-readable medium of claim 9 , wherein the instructions are executable by the at least one processor to cause the system to:
analyze angular velocity data included in the sensor data to compute the first likelihood of collision score.
11 . The non-transitory computer-readable medium of claim 10 , wherein the instructions for causing the system to analyze the angular velocity data are further executable by the at least one processor to cause the system to:
compare the angular velocity data to one or more machine learning datasets corresponding to non-collision events in which a mobile computing device was dropped.
12 . The non-transitory computer-readable medium of claim 9 , wherein the instructions for causing the system to receive the sensor data indicative of the occurrence of the collision are further executable by the at least one processor to cause the system to:
receive sensor data from one or more vehicle based sensors.
13 . The non-transitory computer-readable medium of claim 9 , wherein the instructions for causing the system to receive the sensor data indicative of the occurrence of the collision are further executable by the at least one processor to cause the system to:
receive sensor data from one or more mobile computing devices; and
receive location data indicative of respective locations of the one or more mobile computing devices.
14 . The non-transitory computer-readable medium of claim 13 , wherein the instructions are executable by the at least one processor to cause the system to:
communicate a corroboration request to a first mobile computing device of the one or more mobile computing devices;
after receiving an indication from a user of the first mobile computing device that the collision ocurred, communicate an indication of the collision to an emergency dispatch system.
15 . The non-transitory computer-readable medium of claim 9 , wherein the false positive collection location comprises one or more of: a ski resort, an amusement park, or a body of water.
16 . The non-transitory computer-readable medium of claim 9 , wherein the instructions for causing the system to receive the sensor data indicative of the occurrence of the collision are further executable by the at least one processor to cause the system to:
apply one or more machine learning algorithms to the sensor data, wherein the one or more machine learning algorithms are trained to infer, based on the sensor data, whether a collision occurred;
wherein the instructions are further executable by the at least one processor to cause the system to:
after determining that the first likelihood of collision score does not exceed a predetermined collision threshold, update the one or more machine learning algorithms to indicate that a false positive collision determination was made by the one or more machine learning algorithms.
17 . A computer-implemented method comprising:
after receiving sensor data indicative of an occurrence of a collision and an indication that a data collection location corresponding to the sensor data is within a predetermined radius of a false positive collection location, determining that the collision did not occur;
after receiving an indication that the data collection location corresponding to the sensor data is not within the predetermined radius of the false positive collection location,
determining a first likelihood of collision score;
after determining that the first likelihood of collision score does not exceed a predetermined collision threshold, determining that the collision did not occur; and
after determining that the first likelihood of collision score exceeds the predetermined collision threshold, determining that the collision did occur.
18 . The computer-implemented method of claim 17 , further comprising:
analyzing angular velocity data included in the sensor data to compute the first likelihood of collision score.
19 . The computer-implemented method of claim 18 , wherein analyzing the angular velocity data comprises:
comparing the angular velocity data to one or more machine learning datasets corresponding to non-collision events in which a mobile computing device was dropped.
20 . The computer-implemented method of claim 17 , wherein receiving the sensor data
indicative of the occurrence of the collision comprises:
receiving sensor data from one or more vehicle based sensors.