Class Grouping Levels In Students Using The Davies-Bouldin SOM Index (Self Organizing Map)

Abstract

The data contained in the SMA Wachid Hasyim 2 Taman-Sepanjang is still raw student data so that data needs to be processed. Processing data by grouping (clustering) the data has a variety of methods one of which is the SOM (Self Organizing Map). To validate data after use of distance matrix used IDB (Davies-Bouldin index). IDB in the SOM aims to increase the accuracy of validation of the data analysis. At the end of data analysis with case studies of SMA Wachid Hasyim 2 Taman-Sepanjang  aims to determine the degree of similarity within a group of students (per class). SOM method was able to classify the data adjacent to search based on pattern similarity. The similarity of the data on grouping students as tree  clusters with learning rate 0.6 and epoch 10, 20, 30 with the smallest MSE = 41.42 at epoch 30 on 245 training data. While that made tree until nine cluster with learning rate 0.6 and epoch 10, 20, 30 with the smallest MSE = 25.04 at epoch 20 in cluster 4 with 245 training data. The smallest value in the IDB  cluster validation using tree until nine cluster,  245 training data are in cluster  9  with the value IDB = 74.6 and the results are less accurate because of the class only 2 groups.