Pengelompokkan Ulasan Negatif Pengguna Aplikasi Pedulilindungi Dengan Metode Density-Based Spatial Clustering Of Applications With Noise

Oktaviantono, Aubert (2022) Pengelompokkan Ulasan Negatif Pengguna Aplikasi Pedulilindungi Dengan Metode Density-Based Spatial Clustering Of Applications With Noise. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 06211840000081-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Aplikasi PeduliLindungi diperkenalkan pemerintah sebagai salah satu upaya untuk menekan penyebaran virus COVID-19 dengan dilakukan pendataan, tracing, dan tracking penduduk ketika beraktivitas di luar rumah. Dalam pengembangannya, PeduliLindungi beberapa kali bermasalah ketika digunakan dan pengguna menyampaikan keluhan dan ulasan melalui Google Play Store sebagai platform mengunduh Peduli-Lindungi. Jumlah ulasan yang tidak sedikit dapat menjadi kendala bagi pengembang aplikasi dalam memperoleh kendala yang sering dialami pengguna. Oleh karena itu, penelitian ini bertujuan untuk mengelompokkan ulasan negatif dari pengguna PeduliLindungi menggunakan metode Density-Based Spatial Clustering of Applications with Noise (DBSCAN) dengan metode Genetic Algorithm (GA) sebagai feature selection. Metode DBSCAN diketahui memiliki keunggulan karena tidak memerlukan jumlah initial cluster dan tidak mengelompokkan data outlier dalam klaster manapun. Penggunaan GA dalam seleksi fitur dapat meningkatkan performa pengelompokkan. Selain itu, penelitian ini membandingkan penggunaan metode Term Frequency (TF) dan Term Frequency-Inverse Document Frequency (TF-IDF) sebagai metode ekstraksi fitur teks ulasan, dan n-gram sebagai unit dasar term ulasan, serta dilakukan visualisasi menggunakan word cloud untuk eksplorasi data dan hasil pengelompokkan. Penggunaan kombinasi kata unigram dan bigram berpengaruh terhadap penurunan jumlah noise, sedangkan penggunaan metode TF dan TF-IDF untuk ekstraksi fitur tidak berpengaruh dalam peningkatan nilai silhouette coefficient, namun secara umum TF-IDF memberikan hasil dengan jumlah noise lebih sedikit. Hasil pengelompokkan dari ulasan yang ditujukan pada versi 4.0 terbentuk 2 klaster yaitu klaster kendala sertifikat dan input tanggal lahir, dan klaster ungkapan kekecewaan pengguna. Hasil pengelompokkan dari ulasan pada versi 4.1 terbentuk 5 klaster yaitu klaster kendala sertifikat dan input tanggal lahir, kendala akses sertifikat, klaim sertifikat, server error, dan internal server error.
==============================================================================================================================
The PeduliLindungi application was introduced by the government as one of the efforts to suppress the spread of the COVID-19 virus by collecting data, tracing, and tracking residents when doing activities outside their home. In its development, PeduliLindungi has several problems when used and users submit complaints and reviews through the Google Play Store as a platform to download PeduliLindungi. The number of reviews that are not small can be an obstacle for the app developers in obtaining problems that are often experienced by users. Therefore, this study aims to classify negative reviews from PeduliLindungi users using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) method with the Genetic Algorithm (GA) method as feature selection. The DBSCAN method is known to have advantages because it does not require the number of initial clusters and does not group outlier data into any cluster. Using GA in feature selection can improve clustering performance. In addition, this study compares the use of the Term Frequency (TF) and Term Frequency-Inverse Document Frequency (TF-IDF) methods as a feature extraction method for review texts, and n-gram as the basic unit of review terms, as well as visualization using a word cloud for exploring data and grouping results. The use of a combination of unigram and bigram words has an effect on reducing the amount of noise, while the use of the TF and TF-IDF methods for feature extraction has no effect on increasing the silhouette coefficient value, but in general TF-IDF gives results with less noise. The results of the grouping of reviews aimed at version 4.0 formed 2 clusters, namely the certificate constraint cluster and date of birth input, and the user disappointment expression cluster. The results of the grouping of the reviews on version 4.1 formed 5 clusters, namely the certificate constraint cluster and date of birth input, certificate access constraints, certificate claims, server errors, and internal server errors.

Item Type:	Thesis (Other)
Additional Information:	RSSt 519.53 Okt p-1 2022
Uncontrolled Keywords:	DBSCAN, Ekstraksi Fitur, Genetic Algorithm, Pengelompokkan Teks, Word Cloud. DBSCAN, Feature Extraction, Genetic Algorithm, Text Clustering, Word Cloud.
Subjects:	H Social Sciences > HA Statistics
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Mr. Marsudiyana -
Date Deposited:	11 Jun 2026 03:19
Last Modified:	11 Jun 2026 03:19
URI:	http://repository.its.ac.id/id/eprint/133725

Actions (login required)

View Item