PERANCANGAN MODEL REKOMENDASI PRODUCT BUNDLING BERDASARKAN CUSTOMER SEGMENTATION BANK Z DENGAN MENGGUNAKAN TEKNIK DATA MINING

• Main Author: Kailiffadril, Reyhan
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77672

This research aims to design a product bundling recommendation model based on Bank Z's customer segmentation using data mining techniques. Bank Z is a regional development bank that has been established since 1961. Currently, Bank Z is planning strategic steps to increase bank revenue. Previously, Bank Z experienced a decline in revenue as shown by a decrease in credit growth by its customers. The strategic effort that will be made by Bank Z is to create a product bundling promotion strategy for products and services owned by Bank Z. Until now, the promotional strategy carried out by Bank Z is still generalized for all its customers. This is a problem because the promotions carried out have not been differentiated according to the needs of their customers. After being traced, the root cause of Bank Z's condition is that Bank Z does not have a model that can provide recommendations for promotional strategies according to its customer segments. The methodology used in this research is Cross-Industry Standard Process for Data Mining (CRISP-DM). In determining Bank Z's customer segments, clustering methods are used using partitioning clustering algorithms, namely K-Means clustering and K-Modes clustering. After generating Bank Z’s customer segments, association rule mining with Apriori algorithm is used to determine product bundling recommendations according to Bank Z’s customer segments. This research produces a K-Means clustering model as the best customer segmentation model with an average silhouette index value of 0.21. The model categorizes Bank Z customers into three customer segments. In addition, five recommendations for the best product combinations for each segment that can be used as product bundling strategy product pairs by Bank Z using the Apriori algorithm are also given. After designing the model, an application prototype was developed that can execute the model and display the model results using the Python programming language.