Journal ID : TRKU-11-07-2020-10878
[This article belongs to Volume - 62, Issue - 06]
Total View : 310

Title : Comparative Outlier Detection Design Based on K-Means Clustering for Data Grouping Absorption High School National Exam

Abstract :

Outliers appear as extreme values but often contain information that is very important so it needs to be examined first whether the data is still used or issued. Outlier detection is a hot topic for research. Improved new technologies and a variety of applications cause an increase in the need for outlier detection. The outlier method was successfully applied in various fields, such as education, economics, business, health, space, geology, and credit cards. This study aims to discuss the design of outlier detection methods using the KMeans Clustering method. This study uses four different types of identification tests, there are Extreme Standard Deviation (ESD), Shapiro-Wilk W Test, Dixon-Type Test, and Boxplot-Rule. The results showed that the design of outlier detection methods can be used to compare and find which is more effective in solving problems in the grouping of national exam absorption data in Mathematics. Grouping is done by using 40 indicators of competency achievement on 5 materials tested: Algebra, Geometry and Trigonometry, calculus, and Statistics. Further research can be carried out at the implementation and evaluation stages of the design of this outlier detection method

Full article