Analisis Perbandingan Dan Penerapan Metode K-Nearest Neighbor, Decision Tree, Dan Hybrid KNN-DT Dalam Memprediksi Waktu Pelanggan Berhenti Berlangganan

Khanand, Fuad (2019) Analisis Perbandingan Dan Penerapan Metode K-Nearest Neighbor, Decision Tree, Dan Hybrid KNN-DT Dalam Memprediksi Waktu Pelanggan Berhenti Berlangganan. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 024115400000090-Undergraduated_Thesis.pdf Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Customer relationship management bertanggung jawab terhadap hubungan antara perusahaan dengan pelanggan. Dalam rangka menjalankan fungsinya penting untuk lebih memprioritaskan lamanya pelanggan menjadi pelanggan dibandingkan dengan meningkatkan akuisisi pelanggan baru. Fenomena pelanggan berhenti berlangganan disebut customer churn. Penting bagi perusahaan untuk mengetahui pola perilaku pelanggan untuk mengantisipasi customer churn. Pendekatan metode data mining dapat digunakan untuk memprediksi waktu pelanggan akan berhenti berlangganan dengan pendekatan klasifikasi dengan membagi pelanggan menjadi masing-masing kelasnya bergantung pada bulan ke berapa pelanggan tersebut churn. Tiga metode klasifikasi dilakukan yakni KNearest neighbor, Decision tree C45, dan Hybrid KNN-DT. Parameter pembanding yang digunakan yakni adalah nilai akurasi dan F1-score, dengan menggunakan uji statistik single factor anova dilanjutkan dengan multiple pairwise comparison untuk membuktikan signifikansi perbedaan dari nilai rata-rata parameter. Hasil yang didapatkan yakni bahwa metode decision tree c45 dengan nilai rata-rata akurasi 78,081% dan nilai rata-rata f1 score 75,74% adalah metode terpilih. Uji sistem yang digunakan adalah pada pelanggan Kartu Halo PT Telekomunikasi Seluler area Jawa-Bali Nusra.
=============================================================================================================================
Customer relationship management is responsible for the relationship between the company and the customer. In carrying out its functions, it is important to prioritize the length of time a customer remains subscribed rather than focusing on increasing the acquisition of new customers. The phenomenon of customers unsubscribing is called customer churn. It is important for companies to understand customer behavior patterns in order to anticipate customer churn. A data mining method approach can be used to predict the time a customer will unsubscribe through a classification approach, dividing customers into classes based on the month in which churn occurs. Three classification methods were applied, namely K-Nearest Neighbor, C4.5 Decision Tree, and Hybrid KNN-DT. The comparison parameters used are accuracy and F1-score, evaluated using a single-factor ANOVA statistical test followed by multiple pairwise comparisons to determine the significance of the differences in the mean values of the parameters. The results show that the C4.5 Decision Tree method, with an average accuracy of 78.081% and an average F1-score of 75.74%, was selected as the best-performing method. The test system used is based on Kartu HALO customers of PT Telekomunikasi Selular in the Java-Bali Nusra area.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Data mining, Customer churning time, Customer Relationship Management, K-Nearest neighbor, Decision tree C45
Subjects:	Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) T Technology > T Technology (General) > T174 Technological forecasting
Divisions:	Faculty of Industrial Technology > Industrial Engineering > 26201-(S1) Undergraduate Thesis
Depositing User:	Khanand Fuad
Date Deposited:	22 Jul 2026 03:11
Last Modified:	22 Jul 2026 03:11
URI:	http://repository.its.ac.id/id/eprint/67676

Actions (login required)

View Item