Abstract: Techniques for generating a prediction for a location of a mobile device includes a network controller obtaining signal strength information and angle-of-arrival information for mobile devices from a plurality of access points disposed at a location that includes two or more zones. The network controller extracts features from the obtained information. The features indicate whether the mobile devices are more likely in a first zone than in one or more of second zones of the two or more zones. The network controller generates a software model for classifying a mobile device into one of the two or more zones by applying a computer learning process to the extracted features. The network controller uses the software model to generate a prediction indicating that a first mobile device is in one of the two or more zones, and performs an action at the location based on the prediction.

Abstract: In one embodiment, a device receives location estimates for a wireless node in a network, each location estimate having an associated timestamp. The device applies hierarchical clustering to the received location estimates and their associated timestamps, to identify locations and points in time in which the wireless node was stationary. The device performs sequence modeling on the identified locations and points in time in which the wireless node was stationary, to form a sequence of locations and associated time periods in which the wireless node was stationary. The device associates the wireless node with a behavioral profile based on the sequence of locations and associated time periods in which the wireless node. The device generates, based in part on the behavioral profile for the wireless node, a predictive model that predicts a location of the wireless node at a particular point in time.

Abstract: In one embodiment, a device receives location estimates for a wireless node in a network, each location estimate having an associated timestamp. The device applies hierarchical clustering to the received location estimates and their associated timestamps, to identify locations and points in time in which the wireless node was stationary. The device performs sequence modeling on the identified locations and points in time in which the wireless node was stationary, to form a sequence of locations and associated time periods in which the wireless node was stationary. The device associates the wireless node with a behavioral profile based on the sequence of locations and associated time periods in which the wireless node. The device generates, based in part on the behavioral profile for the wireless node, a predictive model that predicts a location of the wireless node at a particular point in time.