In the present paper, different clustering techniques were applied to detect significant patterns describing single-household water consumption in a residential neighborhood of the City of Naples, basing on hourly time series aggregated at the monthly scale. Comparisons among results were performed by means of a selection of Clustering Validity Indices, that were adjusted to overcome a bias caused by sparsely populated clusters. The most performant cluster solution proved to be the one resulting from the application of a mixed strategy, namely a Self-Organized Map followed by K-means performed on first level cluster centroids.
