Predicting the sweetness level of avomango (Gadung Klonal 21) using multi-predictor local polynomial regression

Abstract

One of the mango’s maturity aspects is the sweetness of the fruits. Mature Avomango has a high degree of sweetness, characterized by a high total soluble solids (TSS) content. Currently, many non-destructive tests are using Near Infra-Red (NIR) spectroscopy to find out the TSS content. NIR spectroscopy generates spectra data, which can be used as predictors to predict Avomangos sweetness level. This study aims to predict the level of Avomangos sweetness by using a multi-predictor local polynomial regression approach and compare it with multiple polynomial regression. In this study, we use 120 samples of Avomango divided into two parts, 100 as training data and 20 as testing data. The multi-predictor local polynomial regression has better performance with the value mean absolute percentage error (MAPE) is 8.554% that categorized as a highly accurate prediction for predicting Avomangos sweetness level.