Pistachio nut varieties sorting by data mining and fuzzy logic classifier

ABSTRACT: In this paper, sorting of two Iranian varieties of pistachio, “Akbari” (Ak) and “Kaleghouchi” (Ka), by the decision tree and fuzzy logic classifier is investigated. 240 Ak and 120 Ka pistachio nuts impacted to a stainless steel plate through a chute, and the acoustic signal of these impacts picked up using a microphone and saved on a computer via a sound card. Further analysis has been made on the statistical parameters extracted from these time-domain signals using J48 algorithm. Feature selection and classifying was carried out, with 200 Ak and 100 Ka as training set as well as 40 Ak and 20 Ka as test set of instances. A correct classification rate (CCR) of 93.33% for training and 96.67% for testing set was observed. The RMSE was obtained 0.24 and 0.23 for training and test sets respectively. Outputs of J48 decision trees (DTs) was converted to IF-THEN rules and fed to the fuzzy inference system (FIS) as inputs, and can be easily utilized by a fuzzy logic controller (FLC) for online sorting applications.