3.1. Rough set analysis
Our explanatory framework of the characteristics for urban green spaces management is
based on a particular recently developed artificial intelligence method viz. rough set analysis.
In recent years, the popularity of artificial intelligence techniques for the identification of
underlying structures in complex databases has drastically risen. In our study, the data system
on urban green spaces can be regarded as a mixed (qualitative and quantitative) database that
is suitable for classification and explanation. This multidimensional classification approach
appears to be able to identify various important factors that are responsible for successes and
failures of urban green space policies.
Rough set analysis is a qualitative multivariate decision-analytical classification method that
originates from the artificial intelligence literature. This method seeks for patterns among
explanatory variables and a relevant ‘endogenous’ variable to be explained. It is based on
mathematical concepts that deal with uncertainty. In the rough set model an upper and lower
bound is defined, each of which has members and non-members respectively. Members of a
boundary region are “possible members”. The classification rules are represented as
“if...then” statements, with the aim to make the maximum reliable prediction for the
assignment of a certain event to a given class.
Rough set data analysis (RSDA) is an application of Knowledge Discovery in Databases
which is concerned with extracting useful information from a complex multivariate data base
(Fayyad et al. 1996). Rough set data analysis is based on minimal model assumptions and
admits ignorance when no proper conclusion can be drawn from the data at hand (Ziarco
1998). RSDA draws all its information from the a priori given data set. In other words,
RSDA remains at the level of an empirical system: more formally, the numerical and the
empirical system coincide and the scaling is the identity function. In RSDA, there is no
numerical system that is different from the operationalisation of the observed data, and there
are no outside parameters to be chosen, nor is there a statistical model to be fitted. RSDA can
be viewed as a preprocessing device to recognize the potentially important explanatory
variables. Data reduction is the main feature of RSDA, as it allows to represent hidden
structures in the database. It should be stressed here that rule induction is not a part of rough
set theory. It can rather be seen as a tool for preparing data for induction especially for
defining classes for which rules are generated. The final outcome of the data base is a