TEDA Regressor: An Unsupervised Regression Technique Based on TinyML Approach
TEDA Regression, TinyML, Unsupervised, Regression, data stream.
The insertion in the network of everyday objects within the reality of the Internet of Things (IoT) brings new possibilities such as data processing at the edge of the application. When processing is performed in a resource-constraints device, such as a microcontroller, and using machine learning techniques, we have the concept of TinyML (Tiny Machine Learning). Within this context, the present work develops an unsupervised regression technique aimed at TinyML and IoT applications involving data streams. The technique is based on the concepts of typicality and eccentricity of samples from the dataset to be processed and uses a recursive least squares filter approach for regression. Preliminary results, extracted from 2 data sets, proved to be promising and with conditions for improvement.