Framework de mineração de dados agropecuários

Abstract

This work presents the development of a framework for data mining of agricultural data. Using the process of knowledge extraction in databases, called Knowledge Discovery in KDD Databases and following its five steps: data selection, pre processing, transformation, mining, validation, and visualization of results. The framework was developed using the Python programming language, using machine learning libraries from scikit-learn and pandas, graph plotting libraries from matplotlib and interface design through the Qt Designer IDE.The K-means and DBScan algorithms were used to perform data grouping, and the decision tree and Naive Bayes algorithms for classification. Tables from the last agricultural censos conducted in 2017, available in the IBGE System for Automatic Recovery-SIDRA, were used to perform a case study presented in section 4 of this work to test the feasibility of the framework. Through the case study, the use of the framework proved to be viable for use by professionals or researchers in the area. The K-means clustering algorithm is equal to 0.78, 0.74, and 0.64 for a K equal to 2, 3, and 4, respectively. The performance of the decision tree-based classifier achieved 100% on the metrics of precision, recall, and f1-score, while the Naive Bayes classifier achieved 80, 96, and 85% for precision, recall, and f1-score, respectively.