Soft Comput (2017) 21:3537–3550
DOI 10.1007/s00500-017-2621-8
FOCUS
Fuzzy transforms prediction in spatial analysis and its application
to demographic balance data
Ferdinando Di Martino
1
· Salvatore Sessa
1,2
Published online: 11 May 2017
© Springer-Verlag Berlin Heidelberg 2017
Abstract We present a new prediction algorithm based on
fuzzy transforms for forecasting problems in spatial analysis.
Our algorithm allows to predict the spatial distribution of
assigned parameters of the problem under exam. Here, we
test our method by exploring the demographical balance data
measured every month in the period 01/01/2003–31/10/2014
in the municipalities of “Cilento and Vallo di Diano” National
Park located in the district of Salerno (Italy). We use this
method for predicting the value of the parameters “birthrate”
and “deathrate” in November 2014. We apply this process
in all the municipalities in the area of study; moreover, we
present a fuzzification process for establishing the thematic
map of the errors calculated between the real data and the
predicted data. The thematic maps are constructed in a GIS
environment.
Keywords Fuzzy transform · Forecasting · GIS · Time
series
Communicated by F. Di Martino, V. Novákc.
B Salvatore Sessa
sessa@unina.it
Ferdinando Di Martino
fdimarti@unina.it
1
Dipartimento di Architettura, Università degli Studi di Napoli
Federico II, Via Toledo 402, 80134 Naples, Italy
2
Centro Interdipartimentale di Ricerca in Urbanistica Alberto
Calza Bini, Università degli Studi di Napoli Federico II, Via
Toledo 402, 80134 Naples, Italy
1 Introduction
In many spatial analysis problems, the decision maker needs
to use support tools that allow to predict the evolution of a
phenomenon by assessing its spatial distribution in a period
of time on the area of study and displaying it on a thematic
map. Prediction models are statistical and applied for estimat-
ing the future trend of a variable starting from a set of numer-
ical data. These models are used in many practical problems
including, e.g., real-time signal processing problems, cur-
rency forecasting, transport planning, land use forecasting,
weather forecasting, etc. Many forecasting methods exist
in literature, as multivariate autoregressive models, nonlin-
ear models, simulation methods, fuzzy interpolation, causal
methods, linear regression: Some comparisons have been
shown in Abraham and Ledolter (1983), Armstrong (2001),
Farnum and Stanton (1989), Makridakis et al. (1998), Palit
and Popovic (1999). Other methods involve soft computing
for supporting uncertainty in the data, as fuzzy rules extrac-
tion algorithms (Cai et al. 2015; Chang and Liu 2008; Chen
et al. 2013; Chen and Hwang 2000; Chen and Wang 1999;
Chissom 1993; Jilani et al. 2008; Lee et al. 2006; Liu 2007;
Palit and Popovic 1999; Shahrabi et al. 2013; Singh and
Borah 2013; Singh 2009; Song and Chissom 1993, 1994;
Sullivan and Woodall 1996; Thawonmas and Abe 1999;
Tsaur et al. 2005; Wang and Mendel 1992), multi-level neural
networks (Aliev et al. 2006; Chen et al. 2006, 2004; Inoue
et al. 2001; Lendasse et al. 2007; Li and Kozma 2003; Li
et al. 2003; Su and Li 2003; Tamura et al. 2008) and genetic
algorithms (EL-Naggar and AL-Rumaih 2005; Li et al. 2006;
Liao and Tsao 2004, 2006).
In Di Martino et al. (2011), a new fuzzy forecasting
method based on fuzzy transforms (for short, F-transforms)
(Perfilieva 2006) is presented. In this method, an optimal
fuzzy partition of the input variable domains is given, but
123