Lailatul Cahyaningrum
Universitas Dian Nuswantoro, Semarang

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

Pengaruh Oversampling dan Cross Validation Pada Model Machine Learning Untuk Sentimen Analisis Kebijakan Luaran Kelulusan Mahasiswa Mufida Rahayu; Ardytha Luthfiarta; Lailatul Cahyaningrum; Alya Nurfaiza Azzahra
JURNAL MEDIA INFORMATIKA BUDIDARMA Vol 8, No 1 (2024): Januari 2024
Publisher : Universitas Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/mib.v8i1.7012

Abstract

The Minister of Education, Culture, Research and Technology issued a new policy on graduation standards for undergraduate and postgraduate students. This policy was delivered on August 29, 2023, on live streaming YouTube Kemendikbudristek at the Merdeka Belajar seminar episode 26: Transformation of National Standards and Higher Education Accreditation. The policy has caused various kinds of positive and negative responses in the community. Based on this problem, this research analyzes the sentiment of how the attitude and response of the community regarding this matter, so that it can be useful for the community in the future. This research uses two algorithms Nave Bayes Classifier (NBC) and K-Nearest Neighbor (KNN) with data collection done through YouTube video comments getting a total dataset of 1085 data. After that, enter the data pre-processing which is then labeled using the Lexicon-based method with the stemming Sastrawi method. Datasets are grouped into positive sentiment and negative sentiment where the labeling results show unbalanced label data. Then the oversampling method Synthetic Minority Over-sampling Technique (SMOTE) is performed so that the data can be balanced and produce good accuracy. The test results after the SMOTE technique show that the NBC algorithm has the highest accuracy compared to KNN. The accuracy results are 74%, precision 74.6%, recall 74% and f1-score 73.9%. While KNN produces an accuracy of 50.2%, precision of 75.2%, recall of 50.2%, and f1-score of 34.5%.