Electronic Nose to Classify Tobacco Origin using a Naive Bayes Method based on Feature Selection

Indonesia is one of the largest tobacco producers in the world, and each region has its unique taste. Indonesian local tobacco is also known to have high quality. As a result, Indonesian tobacco is the most sought-after commodity in the international market. Therefore, this research aims to differentiate the types of tobacco based on their region of origin. The tobacco samples came from the Bantul, Boyolali, and Wonosobo areas. The smell of tobacco will be detected using an electronic nose instrument by applying statistical values as features which include minimum, maximum, average, standard deviation, variance, skewness, and kurtosis values. The seven features obtained are then selected to get the best use of the features. Combinations of all features are examined, from one to seven, in the feature selection process. The results show that the combination of minimum, maximum, average, and skewness features is superior based on the smallest absolute error value. While the classification results by applying these features obtained an accuracy rate of 94.444%, a precision value of 95.238%, and a sensitivity value of 94.444%.