US 11,915,180 B2
Systems and methods for identifying an officer at risk of an adverse event
Nicholas Montgomery, Chicago, IL (US); and Ron Huberman, Chicago, IL (US)
Assigned to Benchmark Solutions, LLC, Chicago, IL (US)
Filed by Benchmark Solutions, LLC, Chicago, IL (US)
Filed on Dec. 23, 2020, as Appl. No. 17/132,458.
Claims priority of provisional application 62/955,812, filed on Dec. 31, 2019.
Prior Publication US 2021/0216927 A1, Jul. 15, 2021
Int. Cl. G06Q 10/02 (2012.01); G06Q 10/0635 (2023.01); G06N 20/00 (2019.01); G06Q 10/0631 (2023.01); G06Q 10/0639 (2023.01); G06Q 50/26 (2012.01)
CPC G06Q 10/0635 (2013.01) [G06N 20/00 (2019.01); G06Q 10/06398 (2013.01); G06Q 10/063114 (2013.01); G06Q 50/26 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A machine-learning based system configured to generate a risk score that predicts whether an officer is at risk of involvement in an adverse event, comprising:
a preprocessing module for collecting and storing historical data about a plurality of characteristics and events relating to a plurality of police officers and creating and storing a first-level residual feature ε=representing the number of arrests that the officer has more or less than the average that the officer's department holding all other variables constant, based on the historical data as follows:
Y=β0+custom characterunit+custom charactergeo+custom charactertime+ε
where:
Y=the number of arrests that have occurred in a certain time period for the officer
β0=the average number of arrests for the officer's department

OG Complex Work Unit Math
ε=the officer's residual=The number of arrests that the individual officer has more or less than the average, holding all other variables constant;
a machine learning module for accessing the stored first-level residual feature and: (a) constructing and storing a plurality of models that predict the risk of an adverse event and generating a risk score, each of the models including the first-level residual feature, (b) running each of the models, with the machine learning module, using the collected, historical data, (c) identifying which one of the plurality of models would have best predicted an adverse event in the past and (d) saving the identified model for use by the system to predict an adverse incident in the future; and,
a model application module for running the identified model using data about the officer and generating: (a) a risk score for the officer and (b) other information relating to the risk score, and for outputting the risk score and the other information relating to the risk score from the model application module to a display for displaying the risk score and information relating to the officer's risk score.