Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Data mining is the non-trivial extraction of implicit knowledge in databases which aims to retrieve useful and new information in a high level of abstraction. The advent of Precision Farming generates databases which, because of their size and complexity, are not efficiently analyzed by traditional methods. The present work aims to test if Data Mining routines are capable to determine the behavior of the yield of a crop as a function of physical-chemical soil properties, in order to allow correction of low yield. Databases were used as object of work, where yield is the meta-attribute and the physical-chemical soil properties are the predictive attributes, obtained through data acquisition on field, in areas of 400 m2. The meta- attribute was obtained by way of a yield map generated by precision agriculture equipment. The data set, with 2388 records were mined using the Decision Tree technique and all values were discerned into two levels. As result, rules were generated finite sets of pairs attribute-value describing models relating yield and physical-chemical soil properties. The confidence of rules was evaluated automatically, making possible the selection of the most qualified ones. By the analysis of the rules by human experts, it was determined that is possible to use the models to determine the behavior of the yield of a crop as a function of physical- chemical soil properties. The developed tool is still without processing acceleration techniques and without techniques of refinement of quality of discovery knowledge, what recalls expectations that the results can be quite improved.