Prediksi Pengguna Churn Pada Aplikasi Mobile Banking Berbasis Graph Machine Learning dengan Framework Topological Relational Inference - Graph Neural Network (TRI-GNN)

Meilani, Zahra Dyah (2023) Prediksi Pengguna Churn Pada Aplikasi Mobile Banking Berbasis Graph Machine Learning dengan Framework Topological Relational Inference - Graph Neural Network (TRI-GNN). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 05111940000069-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (6MB) | Request a copy

Abstract

Saat ini industri perbankan sedang berkembang pesat dan mulai bertransformasi ke arah digital. Meskipun demikian, semakin banyaknya kompetitor dan mudahnya nasabah untuk beralih ke produk perbankan lain menyebabkan rentan terjadinya customer churn. Mayoritas penelitian terkait customer churn di sektor perbankan berusaha mengembangkan model klasifikasi machine learning berdasarkan karakteristik nasabah secara individu. Padahal menurut survei, 71% nasabah meminta pendapat dan rekomendasi ke orang terdekat mereka mengenai produk perbankan. Oleh karena itu, pada penelitian ini dilakukan pembuatan model prediksi customer churn pada aplikasi mobile banking dengan memanfaatkan pembelajaran mesin berdasarkan graf, sehingga hubungan dari setiap pengguna juga dapat dianalisis untuk memprediksi kemungkinan churn. Metode yang digunakan adalah Topological Relational Inference - Graph Neural Network (TRI-GNN), metode ini memanfaatkan analisis topologi pada graf dan mengagregasi fitur yang dimiliki suatu nasabah dengan fitur pengguna-pengguna lain yang memiliki kesamaan topologi dengan pengguna tersebut. Implementasi TRI-GNN untuk membuat model prediksi churn memiliki performa unggul dibandingkan dengan dua metode pembanding. Pada evaluasi dengan dengan data uji, TRI-GNN memiliki F1-score bernilai hingga 75%.
===============================================================================================================================
Currently, banking industry is growing rapidly and is starting to transform towards digital. Nonetheless, the increasing number of competitors and the ease with which customers can switch to other banking products makes it vulnerable to customer churn. Most of the research about customer churn in the banking sector seeks to develop machine learning classification models based on the characteristics of individual customers. In fact, according to the survey, 71% of customers asked for opinions and recommendations from those closest to them regarding banking products. Therefore, in this research, a customer churn prediction model was developed in a mobile banking application by utilizing graph-based machine learning, so that the relationship of each user can also be analyzed to predict the possibility of churn. The method used is Topological Relational Inference - Graph Neural Network (TRI-GNN), this method utilizes topological analysis on graphs and aggregates the features of a customer with the features of other users who have the same topology as that user. The implementation of TRI-GNN to create a churn prediction model has superior performance compared to the two comparison methods. In evaluation with test data, TRI-GNN has an F1-score of up to 75%.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Customer Churn, Graph Convolutional Network, Graph Machine learning, Topology Data Analysis, Topological Relational Inference – Graph Neural Network
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA166 Graph theory Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Zahra Dyah Meilani
Date Deposited:	24 Aug 2023 07:21
Last Modified:	24 Aug 2023 07:21
URI:	http://repository.its.ac.id/id/eprint/102197

Actions (login required)

View Item