Hands-On Deep Learning for IoT

K-nearest neighbor (k-NN) classifier

The k-NN algorithm is a non-parametric method that can be trained using the fingerprinting data coming from IoT devices. This tries to classify the collected RSSI values from the gateways to one of the reference points and not to the coordinates. The input consists of the k-closest RSSI values and the output would be a class membership. An input sample is then classified by a plurality vote of its neighbors, with the object being assigned to the class most common among its k-nearest neighbors.

Technically, if the fingerprinting database consists of (X, y)—with X being the RSSI values and y being the set of already known locations—then k-NN first computes the distance , where x is the unknown sample. Then, ...