Data mining for household water consumption analysis using self- organizing maps

Household water consumption is a part of the human related water cycle that can get into the core of water resources management. Analysis of water consumption data can reveal great potentials of individualized water services planning. Data mining is the process of identifying and extracting potentially useful information from data sets. Self-Organizing Maps (SOMs) is a data mining technique that involves an unsupervised learning method to analyze, cluster, and model various types of large data sets. In this paper, it is presented how the daily water consumption of a household in Sosnowiec, Poland, can be clustered into days of the week, through some features. The features used to discretize the days of water consumption are statistic metrics and time zone consumption metrics. The time zoning is realized in two ways, the first being the typical morning, noon, afternoon, evening and night and the second considering the local working hour time zones of three main working sectors, banks, offices and shops. We use the SOM algorithm in three approaches. In each approach, we use some of the selected features. We have managed to get some clusters with specific features that divide the days of this household in weekdays and weekends.