EvoWave, our first real-world application, deals with the Wi-Fi environment in which a micro-computer is immersed. This environment is defined by the strength of the sig- nal from every Wi-Fi antenna in the neighborhood. It depends especially on available routers and other computers or mobile devices, so it is linked to the context of use of the computer: work, teaching, house, etc. If the Wi-Fi signals from different contexts are dissimilar enough from each other, we expect that Chameleoclust should be able to discriminate these contexts using the data. This corresponds to a dynamic stream problem as new classes, i.e., context of use of the computers, are always susceptible to appear/disappear, and the present Wi-Fi antennas are never the same in different contexts (features appearing and disappearing). This example is also challenging regarding the high dimensionality and noise level of the data. The experimental setup is split in four steps:
The software can be downloaded in the webpage of the EvoEvo Project.
EvoWave in action: confusion matrix and radar charts representing the core-point coordinates