The Design of a Tobacco Quality Detection Tool Using K-NN Classification Method

Abstract

Tobacco is an important commodity on Madura Island, where farmers process it into shredded tobacco sold to middlemen. The quality of the shredded tobacco significantly affects its selling price, determined by factors such as color, aroma, leaf elasticity, and handling. Currently, quality assessment still relies on manual methods dependent on the sensory perceptions of experts. However, the physical and emotional instability of these experts can disrupt the accuracy of these evaluations. To solve this issue, there is a need for a tool that can automatically detect tobacco quality without relying on manual observation. This tool is designed using a series of sensors, including the TCS3200 for color analysis and MQ sensors (MQ4, MQ135, MQ138) for detecting aroma components. The output from these sensors is in the form of ADC values, which are then processed as training data. The K-NN (K-Nearest Neighbors) method will be applied to classify tobacco quality into three categories: low, standard, and good. The distance calculation uses the Manhattan Distance formula with a parameter of K=4, ensuring more accurate results. With this tool, the assessment of tobacco quality becomes more consistent and reliable,because reducing the risk of inaccuracies due to human factors.