Application of SVM and KNN algorithms to build a classification model of makapuno coconuts in Vietnam
Abstract
This paper presents the method and results of classifying gelatinous and non-gelatinous coconuts in Tra Vinh Province, Vietnam. An experimental apparatus is built to sample and process acoustic signals produced from the mechanical impact on sampled coconuts including shaking by hand, knocking by hand, knocking by the machine, and using different materials: stone, plastic, metal. Sound wave signals recorded by the microphone are filtered, extracted for features, trained with labeled data sets, and evaluated as gelatinous and non-gelatinous coconuts. Two algorithms selected and compared are the KNN method (k-Nearest Neighbor) and the SVM method (Support Vector Machine). Experimental results show that the proposed methods are able to accurately classify between gelatinous and non-gelatinous coconuts. In particular, the method of taking samples by hand knocking with plastic rods gives the highest accurate result of more than 90%.
Moraref