Identifikasi Dan Prediksi Tingkat Kematangan Pisang Candi Dengan Fitur Warna Dan Tekstur Menggunakan Metode K-Nearest Neighbor

Abstract

Abstract - Candi bananas is the kind of bananas are fruits that are harvested in raw
conditions. producers that using this bananas for Ingredients, only use their visual
senses to predict the maturity level of bananas. this results in an inaccurate
prediction process and causes production failure. This study aims to build an
application system that can identify and predict the maturity level of candi bananas
by utilizing the color and texture features of bananas. In the construction of the
system, the author uses the Red Green Blue color conversion algorithm to Hue
Saturation Intensity as a color feature and the Gray Level Counseling Matrix
algorithm as the extraction of texture features. From the feature data, for
classification using the K-Nearest Neighbor algorithm to find out the value of the
feature is classified into the mature, half-baked or raw category and find out the
results of predicting how many days the banana will mature. The test results
obtained by using data as many as 380 pieces of candi bananas images with 84%
training data and 16% testing data, results were obtained, the system was able to
identify and predict maturity levels in candi bananas with an accuracy of 90% for
identification of maturity and 73% for predictions the day of maturity of the candi
bananas.