Data mining is useful in finding interesting patterns of hidden information in a database with specified algorithms. Management of uncertainty in the automotive industry supply chain, with case data at PT QQQ that produce car batteries, classification techniques are used to manage uncertainty in the case of engine breakdown. Based on the utilization of classification techniques, performance comparison analysis was carried out from several methods, namely Decision Tree, Bagging, Boosting and Random Forest. The research data is divided into testing data (75%) and training data (25%). This study uses Software R for analysis needs. The need for testing the goodness of the model uses package (caret) help to see the value of accuracy, sensitivity and specificity. The analysis shows that the Random Forest and Bagging method is superior compared to the Decision Tree and Boosting methods based on accuracy criteria, while the sensitivity criteria, Bagging and Boosting methods are superior to Random Forest and DecisionTree. The lowest sensitivity value is owned by the Decision Tree Method, which indicates that the ability of the method is weak in predicting very few classes.
Oleh :
Winnie Septiani
