Relation of home energy consumption and static properties of consumers

Abstract

In the current situation, we need efficient methods to save home energy. Home energy management systems (HEMS) are being now developed, but they require time to change people's style of consumption from the analysis of their behaviour. We analysed a data set with both the amount of consumption and consumers' information, to point out what kind of characteristics of consumers, we called them static properties, would affect the consumption. We use the methodology of machine learning. Here we make a naive Bayes classifier to tell the tendency of consumption from the consumer's static properties. After getting the accuracy of 0.4148, not so high, we estimate the importance of each static property with statistic measure such as χ2 and so on, in order to improve the accuracy and to find the important static properties. Although this process does not bring a significant improvement of the accuracy, we have found several static properties to affect the consumption. We can add a quick diagnostic functionality to the HEMS with these results.