Scholarlink

<p>Customer segmentation was carried out using Recency, Frequency, and Monetary (RFM) values derived from transaction records and grouped using the Bisecting K-Means algorithm. From an initial dataset of 541,909 entries, 397,884 valid records were retained after removing missing values, negative quantities, and duplicate rows. RFM values were calculated and normalized before clustering, and the Elbow method confirmed three as the optimal number of customer groups. The resulting clusters reflected clear behavioral patterns: one group showed high spending and frequent purchases, another included inactive users, and the third represented recent but low-spending customers. The segmentation quality was evaluated using a Silhouette Score of 0.65 and a Davies–Bouldin Index of 0.48. Based on RFM scoring, customers were also categorized into four tiers: Platinum (1263), Gold (1324), Silver (982), and Bronze (770). Compared to standard K-Means, BK-Means produced more balanced and well-separated clusters, making it more suitable for the given dataset.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Intelligent customer segmentation: unveiling consumer patterns with machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月9日

查看全文
  • Naresh Kumar

摘要

Customer segmentation was carried out using Recency, Frequency, and Monetary (RFM) values derived from transaction records and grouped using the Bisecting K-Means algorithm. From an initial dataset of 541,909 entries, 397,884 valid records were retained after removing missing values, negative quantities, and duplicate rows. RFM values were calculated and normalized before clustering, and the Elbow method confirmed three as the optimal number of customer groups. The resulting clusters reflected clear behavioral patterns: one group showed high spending and frequent purchases, another included inactive users, and the third represented recent but low-spending customers. The segmentation quality was evaluated using a Silhouette Score of 0.65 and a Davies–Bouldin Index of 0.48. Based on RFM scoring, customers were also categorized into four tiers: Platinum (1263), Gold (1324), Silver (982), and Bronze (770). Compared to standard K-Means, BK-Means produced more balanced and well-separated clusters, making it more suitable for the given dataset.