DefinePK

Abstract

Classic telemetry systems are usually limited to taking a large variety and number of measurements focused on outdoor air, large concentrations of automobiles and factories, but indoor air pollution has not been addressed with the same intensity. Any of the telemetry techniques generates a large amount of data that implies a great challenge for its analysis; this work demonstrates the application of Machine Learning techniques in telemetry systems focused on the study of indoor air pollution. A telemetry system has been developed which collects data from the environment which are concentrated in a centralized storage unit and are analyzed by automatic learning techniques that can predict the historical behavior of the CO2 concentration based on the variables of the environment, past records of CO2 and independent variables such as time of day, which creates an important tool to detect anomalous behavior in air pollution by CO2. Results of the development of several data prediction models based on Kernel methods to estimate a regression are presented.