Machine learning coordinated wireless networking

A method comprising:

• propagating a small cell wireless mesh network (network) via a plurality of access points communicating via a backhaul channel;
• generating, by the network, a predictive model for movement of devices that connect to the network based on respective historical movements and signal strengths of those devices, wherein the respective historical movements include recorded locations of each of the devices that connect to the network previously and that stopped for at least a threshold period of time;
• and
• in response to connecting by a first access point of the plurality of access points to a first mobile device:
• comparing a current movement of a first mobile device to a past movement of the first mobile device via the predictive model;
• roaming the first mobile device to a second access point of the plurality of access points based on an indication from the predictive model that the second access point has a greatest signal strength for the network to the first mobile device at a location that the predictive model determines the first mobile device will physically remain for a threshold period.

A system comprising:

• a plurality of access points communicating via a backhaul channel that propagate a small cell wireless mesh network (network), the plurality of access points including a first access point and a second access point;
• a memory communicatively coupled with the network and storing a predictive model corresponding to movement of devices that connect to the network based on respective historical movements and signal strengths of those devices, wherein the respective historical movements include recorded locations of each of the devices that connect to the network previously and that stopped for at least a threshold period of time;
• and
• a processor configured to execute instructions stored in the memory that cause the first access point to roam a first mobile device connected thereto to the second access point based on an indication from the predictive model that the second access point has a greatest signal strength for the network to the first mobile device at a location that the predictive model determines the first mobile device will physically remain for a threshold period, the indication based on a comparison of a current movement of a first mobile device to a past movement of the first mobile device via the predictive model.