Pemodelan Multilabel Tweet Media Sosial Mahasiswa untuk Klasifikasi Keluhan

Musthafa, Muhammad Faris (2018) Pemodelan Multilabel Tweet Media Sosial Mahasiswa untuk Klasifikasi Keluhan. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 5113100131-Undergraduate_Theses.pdf]
Preview
Text
5113100131-Undergraduate_Theses.pdf - Accepted Version

Download (1MB) | Preview

Abstract

Pada umumnya sistem informasi akademik di sebuah perguruan tinggi memiliki fitur umum bagi dosen untuk memantau proses perkembangan akademik anak walinya secara aktif. Namun jika dosen wali ataupun orang tua tidak melakukan pantauan secara aktif maka mahasiswa wali yang memiliki permasalahan akademik berisiko drop out dalam proses evaluasi tingkat 1 universitas karena rendahnya pemahaman dosen terhadap mahasiswa walinya. Tujuan dari Tugas Akhir ini adalah membuat rancangan model prediksi keluhan dalam data tweet mahasiswa. Aspek keluhan bisa dibagi menjadi empat kategori: keluhan personal, keluhan subjek, keluhan relasi, dan keluhan institusi.
Metode multilabel yang digunakan adalah Binary Relevance dengan dengan pilihan classifier Naïve Bayes, Simple Logistic, KStar Decision Table, dan Naïve Bayes.
Berdasarkan hasil pengujian pada berbagai classifier Naïve Bayes memiliki performa tertinggi baik dalam aspek akurasi maupun waktu eksekusi. Hasil implementasi sistem Deteksi Multilabel Keluhan menggunakan classifier Naïve Bayes pada delapan puluh data uji terhadap label keluhan personal, subjek, relasi, dan institusi memiliki akurasi masing-masing bernilai 76.47%, 75%, 80%, dan 80%. Hasil deteksi multilabel keluhan yang ditemukan berpotensi digunakan lebih lanjut pada konteks yang lebih luas.
======================================================================================================================== In general, academic information systems in a university have a common feature for lecturers to monitor the process of academic development of their students actively. However, if the lecturers or parent do not actively monitor, the students who have academic problems are at risk of dropping out in the university in level one evaluation process because of the low understanding of the lecturers towards the students. The purpose of this Final Project is to construct detection model of complaint in student tweet data. The complaint aspects can be divided into four categories: personal complaints, subject complaints, relationship complaints, and institutional complaints.
The multilabel method used is Binary Relevance Transformation with classifier choice of Naïve Bayes classifier, Simple Logistic, KStar Decision Table, and Naïve Bayes.
Based on test results on various classifier, Naïve Bayes has the highest performance both in the aspect of accuracy and execution time. The implementation results of the Multilabel Detection of Complaints using Naïve Bayes classifier on eighty test data on personal complaint label, subject, relation, and institution have accuracy of 76.47%, 75%, 80%, and 80% respectively. The results of multilabel detection of found complaints are potentially used further in the wider context.

Item Type: Thesis (Undergraduate)
Additional Information: RSIf 004.678 Mus p
Uncontrolled Keywords: Deteksi keluhan; kegagalan akademik; pemodelan multilabel; media sosial; Early Detection; academic failure; complaint detection modelling, multilabel, social media
Subjects: H Social Sciences > HF Commerce > HF5415.52 Consumer complaints. Complaint letters
Q Science > QA Mathematics > QA279.5 Bayesian statistical decision theory.
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms
Z Bibliography. Library Science. Information Resources > ZA Information resources > Z699.5 Information storage and retrieval systems
Divisions: Faculty of Information and Communication Technology > Information Systems > 57201-(S1) Undergraduate Thesis
Depositing User: Muhammad Faris Musthafa
Date Deposited: 24 Apr 2018 09:01
Last Modified: 14 Jul 2020 06:31
URI: http://repository.its.ac.id/id/eprint/50952

Actions (login required)

View Item View Item