Perbandingan Model Decision Tree, Support Vector Machine dan K-Nearest Neighbors untuk Memprediksi Kualitas Air Minum

Abstract

The need for drinking water is increasing so that appropriate method support is needed to determine water potability. In this study, machine learning models will be implemented including Decision Tree, Support Vector Machine, and K-Nearest Neighbors to determine the best model in classifying drinking water quality from the Kaggle Water Quality dataset. The dataset consists of 3,276 data with 9 parameters consisting of ph, Hardness, Solids, Chloramines, Sulfate, Conductivity, Organic_carbon, Trihalomethanes and Turbidity, and one Potability attribute as a target that indicates the feasibility of consumption. This study will apply several machine learning models consisting of Decision Tree, Support Vector Machine, and K-Nearest Neighbors. Based on the results of the trial using 20% and 30% testing data, the results are close to the same for the confusion matrix model evaluation metrics (Accuracy, F1 Score, Precision and Recall). So it can be concluded that the Decision Tree classification model gets the best Accuracy value among other classification models of 70.50% on 20% testing data and 70.98% on 30% testing data. However, the one chosen as the final classification model is Support Vector Machine because it has the highest value by meeting three requirements with F1 Score, Precision and Recall values of 82.40% each) from the four requirements tested.