Outlier Detection Using K-Means Clustering with Minkowski-Chebyshev distances for Inquiry-Based Learning Results in Students Dataset
Endang Wahyuni, Sugiyarto Surono, Joko Eliyanto · 2021
Outlier appears as an extreme value but often contains very important information, so it is necessary to be studied whether the data remains used or issued. Outlier detection is a hot topic for the study. Increasing new technologies and various applications cause increased requirements of outlier detection. The Outlier method is successfully applied in various fields, namely: economy, business, health, space, geology, and education. Implementation of an outlier of analysis in the education field is often applied to the evaluation of the learning model. Inquire -based learning model is an important component in educational renewal. Learning by this method encourages learners to learn mostly through their active involvement. This study aims to discuss outlier detection by using the K-Means Clustering method on the inquiry-based learning results in students. This study detects outliers with the K-means method using Minkowski-Chebyshev distance. The result of the proposed method will be compared with the extremes of standard deviation (ESD), Box-Plot, and K-Means Clustering using Euclidean distance. The outlier detection results using K-Means Clustering with the Minkowski-Chebyshev and Euclidean distance produce the same result that can detect 3 data as an outlier that is the student with the ID number 7 Exam Value 7.5, ID number 42 Exam Value 9.0, and ID number 72 with the value of 13.5. While the ESD method and Box-Plot are unable to detect any outlier.