Incremental Learning Model for Smart Meters (Mycroft Mind)
The project is done for Mycroft Mind company, which specializes in energy intelligence solutions. We're developing a novel tree-based algorithm for energy demand forecasting that can learn in an online/incremental fashion (without storing historical data on the device), which is optimized for working on edge devices such as smart meters. We've also developed a specialized ensemble using our model as a base learner, providing even better forecasting accuracy while still keeping computational costs comparably low.