PENERAPAN METODE LONG SHORT-TERM MEMORY (LSTM) DAN MULTI-BEHAVIOR RFM UNTUK PREDIKSI CHURN DAN SEGMENTASI PELANGGAN

Nabilah, Annisa Luthfi (2024) PENERAPAN METODE LONG SHORT-TERM MEMORY (LSTM) DAN MULTI-BEHAVIOR RFM UNTUK PREDIKSI CHURN DAN SEGMENTASI PELANGGAN. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002201081-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (15MB) | Request a copy

Abstract

Dalam industri layanan telekomunikasi, perputaran pelanggan adalah masalah yang signifikan karena pendapatan penyedia layanan sangat bergantung pada retensi pelanggan. Dalam persaingan layanan telekomunikasi yang kompetitif ini, penting bagi penyedia layanan untuk mengetahui kepuasan pelanggan terkait fasilitas layanan yang telah diberikan oleh perusahaan. Pembatalan layanan oleh pelanggan atau beralih ke penyedia layanan pesaing biasa disebut churn. Menarik pelanggan baru tidak akan menguntungkan bagi penyedia layanan. Indihome mengalami penurunan jumlah pelanggan pada tahun 2021. Oleh karena itu, dalam penelitian ini, diusulkan pendekatan Multi-Behavior RFM untuk mengelompokkan pelanggan berdasarkan perilaku dan prediksi churn menggunakan model Long Short-Term Memory (LSTM). Berdasarkan hasil evaluasi pada metode Long Short-Term Memory dapat menjadi saran yang membangun untuk perusahaan berdasarkan proses analitik preskriptif dengan memperhatikan fitur-fitur penting yang mempengaruhi prediksi churn pada pelanggan. Tahap selanjutnya akan diberikan rekomendasi strategi untuk mengatasi permasalahan churn pada data yang digunakan. Hasil penelitian yang diperoleh adalah pelanggan dibagi menjadi tiga segmentasi sesuai perolehan skor RFM dari tiga layanan (TV,Internet, dan Telepon) dan data yang tersegmentasi berhasil memperoleh nilai akurasi : 0.9733, presisi : 0.8947, recall : 0.9623, dan f1-score : 0.9274. Dengan pendekatan segmentasi pelanggan berdasarkan perilaku menggunakan Multi-Behavior RFM berpengaruh terhadap hasil performansi model. Terkait rekomendasi strategi perusahaan yang dapat diberikan berdasarkan dua fitur penting tertinggi, yaitu berdasarkan deskripsi produk dan customer item.

In the telecommunication service industry, customer turnover is a significant issue as service providers’ revenues are highly dependent on customer retention. In this competitive telecommunication service competition, it is important for service providers to know customer satisfaction regarding the service facilities that have been provided by the company. Cancellation of services by customers or switching to competing service providers is commonly called churn. Attracting new customers will not be profitable for service providers. Indihome experienced a decrease in the number of customers in 2021. Therefore, in this study, a Multi-Behavior RFM approach is proposed to group customers based on behavior and churn prediction using the Long Short-Term Memory (LSTM) model. Based on the evaluation results of the Long Short-Term Memory method, it can be a constructive suggestion for companies based on the prescriptive analytic process by paying attention to important features that affect customer churn prediction. The next stage will be given strategy recommendations to overcome churn problems in the data used. The research results obtained are customers divided into three segmentations according to the acquisition of RFM scores from three services (TV, Internet, and Telephone) and the segmented data successfully obtained accuracy : 0.9733, precision : 0.8947, recall: 0.9623, and f1-score: 0.9274. With a customer segmentation approach based on behavior using Multi-Behavior RFM affects the performance results of the model. Related company strategy recommendations that can be given based on the two highest important features, namely based on product descriptions and customer items.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Churn, Long Short-Term Memory, Multi-Behavior RFM, Pelanggan, Perilaku, Telekomunikasi. Churn, Long Short-Term Memory, Multi-Behavior RFM, Customer, Behavior, Telecommunications.
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) Q Science > QA Mathematics > QA76.9D338 Data integration
Divisions:	Faculty of Mathematics and Science > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Annisa Luthfi Nabilah
Date Deposited:	08 Aug 2024 01:21
Last Modified:	08 Aug 2024 01:21
URI:	http://repository.its.ac.id/id/eprint/114249

Actions (login required)

View Item