Dry Bean Classification. Analysis from Data Mining

Abstract

The classification of seeds is a fundamental factor in all agricultural processes, especially regarding the obtention of the highest possible profitability of the products once they are ready to be harvested, and to avoid fastidious manual processes. This investigative work performs a data mining analysis, in order to establish the most influential independent variables in the classification of a seed, in this case of a dry bean. The analyzed independent variables are as follows: Area, perimeter, longest axis length, shortest axis length, aspect, eccentricity, convex area, equivalent diameter, extension, solidity, roundness, chromaticity, form factor 1, form factor 2, form factor 3, form factor 4. Likewise, a dependent variable called class is established, which consists of seven states: Barbunya, Bombay, Cali, Dermason, Horoz, Seker and Sira. In this research, the J48 Algorithm of the machine learning and data mining platform called Weka is used, to identify the class to which a seed can belong to, as well as the most influential independent variables in this process. As a result, it was found with an effectiveness greater than 93%, that the most influential variables are: Perimeter, greatest axis length, least axis length.