Resumen:
Background: For agroindustry, crop diseases constitute one of the most common problems
that generate large economic losses and low production quality. On the other hand, recent
research proposes the development of expert systems to solve this problem, making use of
data mining and artificial intelligence techniques. Furthermore, graphs can be used for
storage of different types of variables that are present in an environment of crops, allowing
the application of graph data mining techniques like graph pattern matching. Therefore, the
development of an expert system for crop disease based on graph pattern matching, can
generate a solution for the identification of favorable conditions for a particular disease, as
a starting point for decision-making.
Goals: Develop an expert system based on graph pattern matching to detect favorable
conditions for coffee rust in Colombian crops.
Methods: This work proposes an expert system, characterized from expert knowledge in
coffee rust, as a starting point for the extraction of rules that determine conditions favorable
for this disease, from induction of decision trees, applied to a dataset of monitoring and
cultivation properties. These rules are expressed as patterns of graphs, which are sought
within an information repository crop expressed as graphs, in order to find the similarities of
these patterns, which determine the state of a crop of coffee against rust.
Results: A set of predictive variables for coffee rust, defined from expert’s knowledge; a set
of graph patterns to identify three favorable conditions for rust infection rates; adaptation of
an algorithm for graph pattern matching and an expert system for detecting coffee rust
infection rates, based on graph pattern matching.
Conclusions: Expert knowledge in coffee rust allows the construction of specific predictive
variables for the disease and include it within models generated by data mining techniques.
From these models, can be extracted rules to be expressed as graph patterns, using their
expressiveness and interpretability. Thus, the application of graph pattern matching results
in the condition of a crop against disease. Moreover, the lack of a large amount of data
restricts the quality of model generation process and the system validation.