Implementasi Business Intelligence Menggunakan Aspect Based Sentiment Analysis (ABSA) Dengan Metode Convolutional Neural Network Pada Ulasan Produk Online Marketplace

Wijanarko, Panji Kresno (2021) Implementasi Business Intelligence Menggunakan Aspect Based Sentiment Analysis (ABSA) Dengan Metode Convolutional Neural Network Pada Ulasan Produk Online Marketplace. Masters thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
09211950093003-Master_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2023.

Download (3MB) | Request a copy
[img] Text
09211950093003-Master_Thesis.pdf - Accepted Version
Restricted to Repository staff only

Download (3MB) | Request a copy

Abstract

Platform online marketplace memberikan kebebasan bagi para penggunanya baik penjual dan konsumen untuk membuat konten yang berkaitan dengan informasi produk atau jasa yang diperjualbelikan sebagai contoh konten yang dibuat oleh konsumen dalam platform online marketplace adalah online review. Menurut survei yang diselenggarakan Womply.com terdapat 75% persen dari pelaku bisnis atau perusahaan yang tidak merespon review atau ulasan customer, padahal bisnis yang merespon customernya menunjukkan peningkatan revenue dengan rata rata 4% untuk setiap ulasan customer yang direspon. Analisis sentimen diperlukan untuk mengekstrak infromasi dalam data teks yang jumlahnya banyak untuk mendapatkan sentimen terhadap aspek, layanan atau perusahaan. Penelitian ini bertujuan untuk melakukan klasifikasi sentimen berbasis aspek dengan menggunakan metode Convolutional Neural Network (CNN) dan memberikan rekomendasi manajerial dengan membandingkan best practice dari online marketplace lain untuk masing masing aspek. Dataset yang diperoleh pada Marketplace ‘X’ dipergunakan untuk analisis sentimen berbasis aspek dengan metode CNN. Analsis sentimen berbasis aspek ini menghasilkan nilai yang baik pada nilai akurasi 93.42%, presisi 96.10%, recall 94.56% dan F measure 95.29%. Perumusan strategi diperoleh melalui perbandingan fitur antara Marketplace ‘X’ dengan marketplace kompetitor yaitu Marketplace ‘A’ dan Marketplace ‘B’. Strategi atau rekomendasi yang dihasilkan antara lain untuk meminimalisasi sentimen negatif pada aspek akurasi, kualitas, pelayanan, harga, pengemasan dan pengiriman antara lain membuka akses live streaming untuk semua toko, menambahkan penilai performa dari feedback pembeli, memberikan insentif bagi setiap toko mengupload video pada produk tertentu, menambahkan dashboard persentase chat terbalas, memfasilitasi pembelian kemasan (bubble wrap, kardus), membuat challenge terkait parameter pengukuran performa, menambahkan fitur penilaian terhadap ekspedisi dan membuat fitur rekomendasi berdasarkan dataset yang ada terkait ketepatan pengiriman yang nantinya disesuaikan dengan geografis pembeli.Strategi atau rekomendasi tersebut memiliki impact terhadap marketing cost dan operating cost perusahaan. Dalam oprasionalnya penerapan strategi atau rekomendasi tersebut dilaksanakan oleh Product Management Team, Marketing Team, dan Seller Management Team. ================================================================================================= The online marketplace platform provides freedom for users, both sellers and consumers, to create content related to information on products or services that are traded, as an example of content created by consumers in the online marketplace platform is an online review. According to a survey conducted by Womply.com, there are 75% percent of business people or companies that do not respond to customer reviews or reviews, even though businesses that respond to customers show an increase in revenue with an average of 4% for each customer review that is responded to. Sentiment analysis is needed to extract information in large amounts of text data to get sentiments about aspects, services or companies. This study aims to classify aspect-based sentiment using the Convolutional Neural Network (CNN) method and provide managerial recommendations by comparing best practices from other online marketplaces for each aspect. The dataset obtained on Marketplace 'X' is used for aspect based sentiment analysis using the CNN method. Sentiment analysis based on aspect resulted in good scores at 93.42% accuracy, 96.10% precision, 94.56% recall and 95.29% F measure. The strategy formulation is obtained through a comparison of features between Marketplace 'X' and competitor marketplaces, namely Marketplace 'A' and Marketplace 'B'. The resulting strategies or recommendations include minimizing negative sentiment on aspects of accuracy, quality, service, price, packaging and delivery, including opening live streaming access for all stores, add performance appraisals from buyer feedback, provide incentives for each store to upload videos on certain products, add a reply chat percentage dashboard, facilitating the purchase of packaging (bubble wrap, cardboard), creating challenges related to performance measurement parameters, add an assessment feature for expeditions and making recommendation features based on existing datasets regarding delivery accuracy which will later be adjusted to the buyer's geography. These strategies or recommendations have an impact on the company's marketing and operating costs. In operation, the implementation of the strategy or recommendation is carried out by the Product Management Team, Marketing Team, and Seller Management Team.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Aspek, Convolutional Neural Network (CNN), Online Marketplace, Online Review, Sentimen, Aspects, Sentiment.
Subjects: H Social Sciences > HD Industries. Land use. Labor > HD38.7 Business intelligence. Trade secrets
H Social Sciences > HF Commerce > HF5548.32 Electronic commerce.
T Technology > T Technology (General) > T57.5 Data Processing
Divisions: Faculty of Creative Design and Digital Business (CREABIZ) > Technology Management > 61101-(S2) Master Thesis
Depositing User: Panji Kresno Wijanarko
Date Deposited: 14 Aug 2021 08:33
Last Modified: 14 Aug 2021 08:33
URI: https://repository.its.ac.id/id/eprint/86517

Actions (login required)

View Item View Item