The potential of data exploration methods in identifying the relationship between short-period (daily) water consumption and meteorological factors

Abstract

The purpose of the work was to identify the hidden relationship between water consumption and meteorological factors, using principal component analysis. In addition, clusters of similar days were identified based on relationships identified by k-means. The study was based on data from the city of Toruń (Poland). The analysis was based on daily data from 2014–2017 divided into three groups. Group I included data from the entire period, Group II- from warm half-years (April–September), and Group III-from cold half-years (January–March and October–December). For Groups I and II the extent of water consumption was explained by two principal components. PC1 includes variables that increase water consumption, and PC2 includes variables that lessen water demand. In Group III, water consumption was not linked to any component. The k-means method was used to identify clusters of similar days. In terms of PC1, the most numerous days were Saturdays, and in terms of PC2 Sundays and holidays. It was determined that further research aimed at explaining the specificity of water consumption on particular days of the week is appropriate.