Madyana Patasik, Santi S, Muhardi M, Thabrani R, Husain T

Implementasi Algoritma Support Vector Machine (SVM) pada Pengklasifikasian Sentimen Warganet terhadap Juru Parkir Liar

Introduction

Implementasi algoritma support vector machine (svm) pada pengklasifikasian sentimen warganet terhadap juru parkir liar. Klasifikasi sentimen warganet terhadap juru parkir liar di Makassar menggunakan Algoritma SVM. Model berhasil mengidentifikasi sentimen negatif dengan akurasi 95%, bantu penertiban kota.

Abstract

Juru parkir liar dapat dengan mudah ditemukan di Kota Makassar dan keberadaannya ini sering meresahkan warga. Oleh karena itu, penelitian ini bertujuan untuk mengklasifikasikan sentimen negatif warganet terhadap juru parkir liar tersebut. Dengan menggunakan algoritma Support Vector Machine (SVM), dari 200 data yang dikumpulkan melalui kuesioner daring, 80% (160 responden) digunakan untuk data latih dan 20% (40 responden) untuk data uji. Hasil menunjukkan bahwa model SVM berhasil mengklasifikasikan sentimen, negatif (70% atau 28 responden) dan tidak negatif (30% atau 12 responden) dari 40 data uji dengan tingkat akurasi sebesar 95%, precision  1.00, recall 1.00, dan F1-score 1.00 untuk kelas/label “positif” (sentimen negatif), precision  0.83, recall 0.83, dan F1-score 0.91 untuk kelas/label “negatif” (sentimen tidak negatif). Dengan demikian, dapat disimpulkan bahwa penelitian ini membuktikan efektivitas algoritma SVM dalam mengklasifikasikan sentimen terhadap juru parkir liar. Hasil yang diperoleh dapat menjadi bahan pertimbangan pihak berwenang dalam menertibkan kota, terutama area sekitar pertokoan atau pusat perbelanjaan.

Review

This paper presents a timely study on the application of Support Vector Machine (SVM) for classifying public sentiment regarding illegal parking attendants in Makassar, a prevalent issue that reportedly causes public unrest. The research addresses a relevant social problem by attempting to gauge and categorize netizen sentiment, which could inform urban policy and regulation. Utilizing an online questionnaire to gather 200 data points, the authors leverage SVM to categorize sentiment into "negative" and "non-negative," demonstrating a practical approach to applying machine learning in local governance contexts. The methodology involves training an SVM model on 160 samples and testing on 40. The reported performance metrics are notably high: an overall accuracy of 95%, with precision, recall, and F1-score all reaching 1.00 for the "positive" class (which is confusingly defined as "negative sentiment"). For the "negative" class (defined as "non-negative sentiment"), the metrics are also strong at 0.83 precision, 0.83 recall, and 0.91 F1-score. While these results suggest excellent model performance, the small dataset size, particularly for the test set (40 samples), raises questions about the generalizability and robustness of these findings. Furthermore, the abstract's inverted labeling convention ("positif" meaning "negatif" sentiment) is confusing and should be clarified for better readability and standard practice. The collection of "warganet" sentiment via a questionnaire, rather than direct social media scraping, also implies a specific type of data that might differ from organic netizen discourse, which should be acknowledged. To enhance the study's impact and validity, several aspects could be considered for future work. Firstly, expanding the dataset size significantly, potentially by incorporating actual social media data, would provide a more representative and robust foundation for classifying "warganet" sentiment and solidify the model's generalizability. Secondly, a clearer and more conventional labeling scheme should be adopted. Thirdly, comparing the SVM's performance against other established sentiment analysis techniques, such as Naive Bayes, Logistic Regression, or deep learning models, particularly for text data, would provide valuable benchmarks. Finally, a detailed discussion on the nature of the data collected via the online questionnaire—whether it involved free-text responses or pre-defined sentiment scales—would offer crucial context to the SVM's classification task and the interpretation of the impressive results. Despite these considerations, the research offers a promising initial step towards leveraging AI for urban management and public feedback analysis.

Full Text

Comments

You need to be logged in to post a comment.