Classification of Salt Quality based on Salt-Forming Composition using Random Forest

Abstract

Abstract Salt is part of the chemical that can be used and needed by humans in the field of consumption or industry. The formation of salt can be done in several ways, namely with seawater or lake water that is evaporated to produce salt crystals or through the process of mining rock salt. The results of the salt obtained will have a different composition depending on the process of formation, the difference in composition can affect the quality of the salt produced, so not all salt results are suitable for consumption. Generally, the salt quality classification process is still done manually, but this method takes a long time and is less effective. So, to overcome this problem, this research utilizes data mining science in classifying salt quality automatically using the Machine Learning algorithm, namely Random Forest. The data used in this study is a salt dataset with 7 attributes and 4 target classes totaling 349 data where the data is divided into training data and test data using k-fold cross validation with different k-fold values, namely 5-fold, 10-fold, and 20-fold. folds. The test results obtained indicate that the value of k = 10 has the best performance by achieving an AUC value of 96.1%, then for the classification accuracy is 87.7%, f1 score is 87.6%, precision is 87.7% and recall is 87.7%.