Analisis Sentimen Berbasis Aspek Pada Ulasan Pelanggan Menggunakan Transformer Neural Network Untuk Optimasi Layanan Retail Healthcare

Iskandar, Dipa Anasta (2025) Analisis Sentimen Berbasis Aspek Pada Ulasan Pelanggan Menggunakan Transformer Neural Network Untuk Optimasi Layanan Retail Healthcare. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032231085-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Digitalisasi memaksa industri retail beradaptasi dengan membuat sebuah strategi saluran distribusi yaitu multi-channel dan omni-channel. Namun, penerapan strategi ini memiliki beberapa hambatan seperti pengintegrasian sistem yang kurang akan membuat pelanggan atau penggunanya kebingungan, dan bahkan jika terlalu berlebihan dalam mengintegrasikan sistem akan terjadi over expense. Analisis yang dilakukan pada penelitian ini bersifat deskriptif dan prediktif pada input data ulasan yang diharapkan menghasilkan arah pengembangan layanan retail yang optimal. Digunakan metode web scraping untuk memperoleh data ulasan pelanggan yang tersedia pada platform Google Play Store untuk perusahaan healthcare yang bergerak pada industri retail dengan strategi multi-channel dan omni-channel. Data yang diperoleh tidak memiliki label aspek dan sentimen untuk dianalis. Untuk mendapatkan label sentimen, teknik Query by Committee digunakan. Untuk mendapatkan label aspek, Topic Modeling dengan BERTopic digunakan. Model klasifikasi dengan basis XLM-RoBERTa dibentuk untuk memprediksi sentimen. Karena luasnya aspek, BERTopic kembali digunakan untuk analisis sentimen berbasis aspek pada setiap label sentimen dan label aspek yang sebelumnya telah diperoleh. Hasilnya, model klasifikasi XLM-RoBERTa memiliki Accuracy 95%, Precision 95%, Recall 95%, dan F1 Score 95%, dengan AUC Score pada label negatif 0.99, label netral 0.95, dan label positif 1.00 pada data uji. Hasil analisis pada penelitian ini memberikan rekomendasi yang komprehensif pada setiap aspek-sentimen, yaitu terkait Convenience, Delivery, Helpful, Integration, Service, Price & Promotion, Technical, dan Other adalah aspek yang ditemukan pada data penelitian ini.
=====================================================================================================================================
Digitalization forces the retail industry to adapt by creating a distribution channel strategy, namely multi-channel and omni-channel. However, the implementation of this strategy has several obstacles, such as insufficient system integration, which can confuse customers or users. Additionally, excessive system integration can lead to over-expenditure. The analysis conducted in this study is descriptive and predictive, based on review data input, aiming to produce optimal retail service development directions.Web scraping methods were used to obtain customer review data available on the Google Play Store platforms for healthcare companies operating in the retail industry with multi-channel and omni-channel strategies. The data obtained did not have aspect and sentiment labels to analyze. To obtain sentiment labels, Query by Committee technique was used. To obtain aspect labels, Topic Modeling with BERTopic was used. A classification model based on XLM-RoBERTa is formed to predict sentiment. Due to the breadth of aspects, BERTopic is again used for aspect-based sentiment analysis on each sentiment label and aspect label obtained. As a result, the XLM-RoBERTa model built for sentiment classification has 95% Accuracy, 95% Precision, 95% Recall, and 95% F1 Score, with AUC Score on negative labels 0.99, neutral labels 0.95, and positive labels 1.00 on test data. The results of the analysis in this study provide comprehensive recommendations on each aspect-sentiment, namely related to Convenience, Delivery, Helpful, Integration, Service, Price & Promotion, Technical, and Other are aspects found in this research data.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Retail, Omni-channel, Multi-channel, Layanan, Transformers, Neural Network, Topic Modeling Retail, Omni-channel, Multi-channel, Services, Transformers, Neural Network, Topic Modeling
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T59.7 Human-machine systems.
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Dipa Anasta Iskandar
Date Deposited:	17 Jun 2025 07:49
Last Modified:	17 Jun 2025 07:49
URI:	http://repository.its.ac.id/id/eprint/119194

Actions (login required)

View Item