Model Analisis Classification dengan J48 untuk Data Mahasiswa dan Dosen di Perguruan Tinggi

Ayub, Mewati and Caroline, Maresha and Christian, Tjio Marvin (2014) Model Analisis Classification dengan J48 untuk Data Mahasiswa dan Dosen di Perguruan Tinggi. In: Seminar Nasional Teknologi Informasi dan Multimedia, 9 Oktober 2014, Surabaya.

[img]
Preview
Text
Model Analisis Classification dengan J48.pdf - Published Version

Download (7Mb) | Preview

Abstract

The existence of historical data in universities, for instance, the data of lecturer and student academic process, is very valuable because of the fact that historical data can be analyzed to extract the implicit knowledge with the use of various analysis methods, which in turn can be used as a basis for improving education system. This research is aimed to analyze the historical data of students and lecturers using predictive classification methods as a continuation from previous research that had been managed to produce a data warehouse schema for both types of data. In addition, the research is focused on three datasets, which are datasets of research and community services and also the dataset of graduates that were extracted using the classification methods of decision tree, especially J48 with some confidence factor parameter settings as well as several minimum numbers of instances for the leaves that will produce optimal analysis model. Research methodology began with the formation of datasets derived from data star schema of the previous research results, followed by the determination of attributes of the datasets that would be used as training data and test data. Furthermore, those datasets were analyzed using some classification models before they finaly were evaluated. The results of some test cases indicated that the decreasing in the value of the confidence factor and also the increasing of minimum number of instances on leaves in J48 classification, both were affect the resulting three pruning. For research and community services datasets, the convergence on the number of leaves in the tree of the decrease of confidence factor value. Meanwhile, for the graduate dataset, in cases of different classes, the convergence of the leaves number was influenced by the distribution of the data in the class attribute, either by adding the minimum value of the instance on the leaves as well as by lowering the confidence factor value.

Item Type: Conference or Workshop Item (Paper)
Uncontrolled Keywords: data mining, classification, decision tree, J48, universities historical data
Subjects: T Technology > T Technology (General)
Depositing User: Perpustakaan Maranatha
Date Deposited: 20 Apr 2015 05:38
Last Modified: 20 Apr 2015 05:38
URI: http://repository.maranatha.edu/id/eprint/11896

Actions (login required)

View Item View Item