Keywords

water supply systems; water quality analysis; self-organizing maps; k-means clustering

Location

Session C1: VI Data Mining for Environmental Sciences Session

Start Date

13-7-2016 3:50 PM

End Date

13-7-2016 4:10 PM

Abstract

Universal access to drinkable water is a constitutional right guaranteed in Brazil. However, not all cities in this country are able to supply the population with the expected quality. Important actions should be taken to improve the supply of drinkable water. To define a strategic plan for this purpose, classification tools can facilitate the design of plans, by grouping cities with similar quality conditions. During the last decades, neural network approaches have been used in environmental models, allowing more accurate representation of some complex systems. This work proposes the use of self-organizing maps (SOM’s) coupled with the k-means algorithm to determine city groups (clusters) based on water quality features available at the National System of Sanitation Information (SNIS). Using the Calinski-Harabaz (CH) index for clustering performance analysis, an optimal number of clusters is defined. The objective of this clustering is to clarify the real conditions, to understand the main service deficits from the water quality perspective, and to plan suitable strategies to reduce these deficits.

 
Jul 13th, 3:50 PM Jul 13th, 4:10 PM

Water Supply System Classification for Water Quality Improvement

Session C1: VI Data Mining for Environmental Sciences Session

Universal access to drinkable water is a constitutional right guaranteed in Brazil. However, not all cities in this country are able to supply the population with the expected quality. Important actions should be taken to improve the supply of drinkable water. To define a strategic plan for this purpose, classification tools can facilitate the design of plans, by grouping cities with similar quality conditions. During the last decades, neural network approaches have been used in environmental models, allowing more accurate representation of some complex systems. This work proposes the use of self-organizing maps (SOM’s) coupled with the k-means algorithm to determine city groups (clusters) based on water quality features available at the National System of Sanitation Information (SNIS). Using the Calinski-Harabaz (CH) index for clustering performance analysis, an optimal number of clusters is defined. The objective of this clustering is to clarify the real conditions, to understand the main service deficits from the water quality perspective, and to plan suitable strategies to reduce these deficits.