Machine-Learning Assisted Prediction of Light Consumption

Abstract

IoT devices enable an assortment of large amounts of information from the environment that may well be used to produce sensible autonomous buildings. For example, within the context of public buildings like hospitals, it's necessary to regulate and predict the indoor environment. Just in case of emergencies or failures, the facility is going to be able to automatically regulate the indoor parameters by using models that learn from the previous usage. This can allow us to acquire advantage of the IoT sensor data and to understand real-time control of the building. The information stream will be unendingly processed and analyzed on-premises. Anomalies can be monitored, and alarms can be triggered accordingly. The objective of this research paper is to construct a classifier to predict the usage of light from the IoT information stream. One manner might be to implement a cloud-based device where all IoT sensors send measurements to an endpoint on the cloud for storage and training of the model. Proposed model will process the IoT data stream and construct a classifier that will be able to accurately predict the target label. The target label for use case is “light.” The research study goal is to suit a model that may predict if the light is on or off and to determine the accuracy of the predicted model using Hoeffding and Naïve Bayes classifier.