Smart Light Recommending System Using Artificial Neural Network Algorithm

This paper proposes smart light recommending system based on sleep monitoring data using Artificial Neural Network (ANN) algorithm. lights are one of the biggest contributors to the power consumption of electrical equipment. The effort to reduce the use of lights is related to not using them when they are not in use. One of the moments when the lights are not used is when we are sleeping but some user sleep styles keep the lights on even when they are sleeping so the idea is how to turn off the lights after the user sleeps. To realize this idea, it is necessary to detect sleep which can be derived from the detection of heart rate by certain sensors. The sensors that are needed are those already embedded in Fitbit, a bracelet-shaped device that can detect various kinds of conditions in the human body. However, Fitbit cannot directly provide the sleep condition to the user or the lamp, but it can provide information in the form of logs, so that the lamp settings cannot be instantaneous. To overcome this problem a learning method can be applied to know sleep patterns that appear in logs produced by Fitbit. This paper applies ANN back propagation to learn the sleep patterns of users, especially sleep start time and sleep end time. Nine ANN models made from user sleep data are applied to 60 days of testing. From these models, the best results were given by the model which gave 82.27% accuracy for sleep start time and 98.28% for sleep end time. Accuracy is largely determined by the user's sleep style.