Comparative Performance Evaluation Results of Classification Algorithm in Data Mining to Identify Types of Glass Based on Refractive Index and It’s Elements

Abstract

Data science is becoming familiar to the public and companies in the era of the Industrial Revolution 4.0. One part of data science is data mining. Data mining is the process of collecting information to see patterns from very large datasets and data discovery which is processed in such a way as to become knowledge based on the interpretation of the information obtained. This paper aims to compare the performance evaluation results of several classification algorithms in data mining (such as DT C-45, Neural Network, KNN, LDA, Naïve Bayes, SVM, and Rule Induction) for identifying types of glass based on its elements and the refractive index. The dataset used is a glass identification dataset from the UCI Machine Learning Repository. The results of the evaluation can be seen from the criteria like Accuracy and Kappa using 10-fold-cross validation. As a result, the K-Nearest Neighbors (KNN) algorithm has the best Accuracy and Kappa values, namely 72.90% for Accuracy and 0.632 for Kappa values. To determine the significance of the accuracy value, the T-Test method is used.