Geoscience Reference
In-Depth Information
of applications and found that only 10% of such systems were actually successful. This disappoint-
ing picture of ES was formulated in the 1990s after the initial interest in these systems began to
wane. An inherent weakness of ES is the fact that they are based on
shallow knowledge
, that is,
empirical and heuristic knowledge rather than
deep knowledge
based on the structure, function
and behaviour of objects. The system does not really have an understanding of causes and effects
and can therefore not self-adapt through a higher process of reasoning, reaching what is inherently
its natural cognitive limits (Turban 1995; Giarratano and Riley 2005). In addition, the process of
knowledge acquisition, that is, the gathering and transferring of human knowledge into the system,
is considered to be the major obstacle in building an ES since either the knowledge is not always
readily available or it can be difficult, expensive and time-consuming to extract, which is referred
to as the
knowledge acquisition bottleneck
(Moore 2000). Furthermore, ES tend to be focussed on
narrow problem domains so their applicability and durability are limited to small and well-defined
real-world problems (Rodriguez-Bachiller and Glasson 2004). Despite these criticisms, ES applica-
tions still continue to appear in the literature, either as separate collections (e.g. Tyler 2007) or in
relevant international journals such as
ES with Applications, Knowledge-Based Systems
and the
International Journal of Knowledge-Based and Intelligent Engineering Systems
. The wide-ranging
applications that appear regularly in these journals clearly indicate that the field of ES is strong
across many disciplines and continues to grow.
ES can be divided into two broad categories, that is, conventional and fuzzy logic-based sys-
tems (Aly and Vrana 2006; Patel et al. 2012). Conventional ES are mainly based on classical or
Boolean logic using symbolic reasoning engines, while a fuzzy ES uses a collection of fuzzy
membership functions and rules, which is capable of handling uncertain or imprecise information
in order to reason about the data. For example, a rule within a fuzzy logic-based ES could be the
following: if
x
is high and
y
is low, then
z
is medium, where
x
and
y
are input variables,
z
is an
output variable and low, medium and high are membership functions defined for variables
x
,
y
and
z
, resp e ct ively.
ES can also be characterised by the scope of the system, for example, real-time ES, web-based
ES, visual ES and spatial ES. The purpose of this chapter is to focus specifically on spatial ES, which
is one approach for problem-solving within the broader set of tools that comprise GeoComputation
(GC). We start by distinguishing between ES and more conventional computer programs and then
provide a checklist for determining when to use an ES for problem-solving. This is followed by an
outline of the basic components of an ES and a description of the basic development methodology
and tools available for building ES. A review of the integration of GIS with ES is then provided fol-
lowed by relevant interoperability mechanisms and a suggestion for an ideal spatial ES framework.
An example of a recently developed ES is then described to illustrate the application of ES to a
geospatial problem ending with an outlook of ES in the future.
11.2 CONVENTIONAL SYSTEMS VERSUS ES
ES differ from other types of computer programs or more conventional systems. Moore (2000,
originally cited in Dantzler and Scheerer, 1993) provides a list of the main differences, which can
be distinguished by the following four criteria: goals, focus, approach and output. The goal of a con-
ventional system is to run algorithms to carry out specific tasks, for example, calculate an equation
or search for an optimal parameter. The goal of an ES, on the other hand, is to encapsulate expert
knowledge and then provide others with access to this knowledge, for example, deciding on where
to mine for minerals. Another difference between conventional systems and ES is their primary
focus, where conventional systems are focussed on data while an ES is focussed on knowledge. The
structures for storing this information therefore require different representations. Relational data-
bases may work well for storing data, but other mechanisms such as object-orientated or logic-based
approaches are better suited to storing knowledge. The approach taken by conventional systems
