Aplikasi Penentu Subyek Skripsi Menggunakan Metode Support Vector Machine

Abstract

Thesis is a task given by the university to students as a final assessment of the learning process that has been taken for several semesters. After completing the thesis, students submit their research results to the campus as a thesis collection. At Petra Christian University, every thesis collected is given a subject as the thesis category. However, giving this subject is still manual, so we need a system that can help determine the subject of the thesis.

The system that is equipped with text mining features will help the library in determining the subject of the thesis. The steps taken are preprocessing consisting of punctual removal, stopword removal, and stemming. Then the process of extracting text data into numbers using TF-IDF. Furthermore, the data will be trained using the Support Vector Machine method which will produce a model and be used to predict subjects from input text. The trained data is the title data and abstract of the existing thesis..

The results of the research conducted showed that in the construction of the SVM classification model the parameters TF-IDF max_df 1, n-gram (1,2), smooth_idf and sublinear_tf true, linear SVM kernel with C 100 for the thesis title and max_df 0.25, n-gram (1,1), smooth_idf and sublinear_tf false, rbf SVM kernel with C 100 and gamma 0.01 for the thesis abstract. Both the title and abstract of the thesis require preprocessing, resample, and l2 normalization.