Device, system, and method for training machine learning algorithms to generate talkgroups based on in-call data

A method comprising:

• generating, at a computing device, via one or more machine-learning algorithms, an initial talkgroup based on a public-safety incident indication of an incident, the initial talkgroup comprising communication devices that communicate on one or more given channels, the one or more machine-learning algorithms initially trained to:
• generate talkgroups based on public-safety incident indications;
• and make changes to the talkgroups;
• receiving, at the computing device, in-call data comprising one or more of communications in the initial talkgroup and call-transmission metadata;
• generating, at the computing device, via the one or more machine-learning algorithms, a suggested change to the initial talkgroup based on the in-call data;
• determining, at the computing device, a machine-learning score of the initial talkgroup or the suggested change, the machine-learning score indicative of a threshold-based positive reinforcement or a threshold-based negative reinforcement of efficiency of the initial talkgroup or the suggested change, wherein the threshold-based positive reinforcement and the threshold-based negative reinforcement of efficiency are respectively indicative of efficient, or inefficient, use of network resources;
• and
• implementing, at the computing device, a machine-learning feedback loop that includes providing the machine-learning score to the one or more machine-learning algorithms to further train the one or more machine-learning algorithms,
• wherein the in-call data comprises one or more keywords identified in the communications of the initial talkgroup, and
• wherein the machine-learning score comprises the threshold-based positive reinforcement of efficiency of the initial talkgroup or the suggested change when the one or more keywords includes one or more of a first set of positive keywords, and
• wherein the machine-learning score comprises the threshold-based negative reinforcement of efficiency of the initial talkgroup or the suggested change when the one or more keywords includes one or more of a second set of negative keywords.