This paper was designed to deal with a real problem of statistical analysis in Geographic Information System (GIS) field. The setting up was the analysis of the Almeria GIS data matrix consisting of 376 instances, designated as all the irrigation communities in Almeria (Spain) that are grouped geographically by 18 water management areas and technically by 38 distinct water sources patterns. The data, collected by Violeta Cabello, belonging to a set of pilot case studies approached during the first year of the H2020 MAGIC project (G.A. n. 6896669). The real GIS dataset was exploited as a test for the fruitful interaction of domain experts and statisticians. The main purpose is to find a partition of irrigation communities such to predict the water consumption for each area. Within the framework of recursive partitioning algorithms by tree-based methods proposed by Breiman at al. (1984), we propose to build an explorative Regression Tree Geographically Weighted. Classification And Regression Trees (CART) can be considered as the ancestors of supervised statistical learning paradigm introduced by Vapnik (2013). Tree-based methods have been proposed for both prediction and exploratory purposes. Since the entire population is surveyed, we explore the irrigation communities of Almeria based on water consumption per hectare. Exploratory trees belong to data mining methods where an important role is the visualization of the results because of it helps the analyst to better understand the phenomena under study (Fayyad et al., 2002). The aim is classifying the specific consumption of water (per hectare) of the farming communities based on either water management areas and different mix of sources for irrigation water (surface, groundwater, waste water, desalination). Even if the regression tree is a device simple to be presented and understood, main issue is the tailoring of the standard approach (being the observations weighted geographically). The results show that the water consumption per hectare in some management areas is quite more spread than in some others. Moreover, the role of profiles of water sources varies in different sets of water management areas, highlighting as the two predictors interact. Finally, this proposal reducing the gap between theory and practice (domain experts) points out the usefulness of our contribution in the knowledge discovery process from databases and collaborative quantitative story-telling for MAGIC project addresses for informing nexus security.

Spatially Weighted Exploratory Regression Tree: A Quantitative Story-Telling Proposal / Iorio, Carmela; Pandolfo, Giuseppe; Siciliano, Roberta. - (2019), pp. 21-22. (Intervento presentato al convegno European European Conference on Data Analysis (ECDA) 2019 tenutosi a Universitat Bayreuth nel 18 - 20 march 2019).

Spatially Weighted Exploratory Regression Tree: A Quantitative Story-Telling Proposal

Carmela Iorio
;
Giuseppe Pandolfo;Roberta Siciliano
2019

Abstract

This paper was designed to deal with a real problem of statistical analysis in Geographic Information System (GIS) field. The setting up was the analysis of the Almeria GIS data matrix consisting of 376 instances, designated as all the irrigation communities in Almeria (Spain) that are grouped geographically by 18 water management areas and technically by 38 distinct water sources patterns. The data, collected by Violeta Cabello, belonging to a set of pilot case studies approached during the first year of the H2020 MAGIC project (G.A. n. 6896669). The real GIS dataset was exploited as a test for the fruitful interaction of domain experts and statisticians. The main purpose is to find a partition of irrigation communities such to predict the water consumption for each area. Within the framework of recursive partitioning algorithms by tree-based methods proposed by Breiman at al. (1984), we propose to build an explorative Regression Tree Geographically Weighted. Classification And Regression Trees (CART) can be considered as the ancestors of supervised statistical learning paradigm introduced by Vapnik (2013). Tree-based methods have been proposed for both prediction and exploratory purposes. Since the entire population is surveyed, we explore the irrigation communities of Almeria based on water consumption per hectare. Exploratory trees belong to data mining methods where an important role is the visualization of the results because of it helps the analyst to better understand the phenomena under study (Fayyad et al., 2002). The aim is classifying the specific consumption of water (per hectare) of the farming communities based on either water management areas and different mix of sources for irrigation water (surface, groundwater, waste water, desalination). Even if the regression tree is a device simple to be presented and understood, main issue is the tailoring of the standard approach (being the observations weighted geographically). The results show that the water consumption per hectare in some management areas is quite more spread than in some others. Moreover, the role of profiles of water sources varies in different sets of water management areas, highlighting as the two predictors interact. Finally, this proposal reducing the gap between theory and practice (domain experts) points out the usefulness of our contribution in the knowledge discovery process from databases and collaborative quantitative story-telling for MAGIC project addresses for informing nexus security.
2019
Spatially Weighted Exploratory Regression Tree: A Quantitative Story-Telling Proposal / Iorio, Carmela; Pandolfo, Giuseppe; Siciliano, Roberta. - (2019), pp. 21-22. (Intervento presentato al convegno European European Conference on Data Analysis (ECDA) 2019 tenutosi a Universitat Bayreuth nel 18 - 20 march 2019).
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11588/749418
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact