Thermostat data mining

Authors

Doc. Ing. Radim Burget, Ph.D., Ing. Lukáš Povoda, Ing. Martin Rajnoha, Ing. Jan Mašek, Ph.D., Ing. Václav Uher

Download
Software is an exclusieve property of Honeywell company

Popis
This project was related to data analysis of thermostats, which are stored to thermostat clouds. Currently there is for about 1.5 million thermostats in cloud which reports 2TB of logs each month about their behavior and indoor and outdoor temperature. The overall history of the data is approximately two years. Source code, which is a part of this handover document is devoted to preprocess the data, analyse the data and visualize reports about the data. Data preprocessing transforms raw log records of thermostat sensors into a form that is suitable for the data analysis. Analysis part uses machine learning in order to predict energy consumption according to settings of thermostat devices. Visualization is an web server which is devoted to provide user interface to the end user. User can change settings of the thermostats and estimate future consumption percentage increase or decrease according to the current settings.

Obrázek1

Projects
Software is an outcome of project HS18657087.

License
Software is an exclusieve property of Honeywell company