Dager: uma ferramenta computacional para agrupamentos em mineração de dados agrícolas georreferenciados

Abstract

Agriculture demands for various computing solutions, especially when refers to the Precision Agriculture (PA). Data Mining is one of the computing resources that can benefit the analysis of data from AP, which usually are georeferenced. However, there are limitations of algorithms and computational tools when it is need to group different characteristics also considering its geographical position. In this context, the aim of this work was to develop and implement algorithms that consider clustering and visualization of georeferenced attributes along with various attributes in the agricutural database. It was created a new computational tool for georeferenced agricultural data mining called Dager. The algorithms PAM, CLARA and CLARANS were implemented and, based on these two new algorithms, and GCLARA GCLARANS were developed and implemented in the tool. Besides the algorithms it was implemented a module for graphical visualization of clusters. For the experiments, a database obtained by Precision Farming and evaluation groups were employed statistical methods ANOVA and MANOVA. The result showed the mapping and visualization of regions within a field with similar characteristics, achieving the proposoal objectives.