Enhanced Customer Segmentation Using LRFSM Behavior Model and Multi-Algorithm Clustering Approaches

Abstract

Customer segmentation plays a critical role in optimizing personalized marketing and driving customer engagement in online retail. Building on the previously established LRFS (Length, Recency, Frequency, Spend) model, this study introduces an enhanced LRFSM framework by incorporating Monetary value as a fifth behavioral dimension. This expansion enables a deeper analysis of customer purchasing behavior and economic contribution. By applying advanced segmentation techniques and evaluating cluster quality through both internal validation scores and external classification metrics, this work ensures both statistical rigor and business relevance. Detailed profiling of each customer segment further provides meaningful insights into targeted strategy development. The results demonstrate that the inclusion of Monetary value not only improves segmentation precision but also supports the creation of scalable, adaptive models tailored to dynamic retail environments, bridging the gap between data science and actionable business outcomes.

Introduction

• With the rise of e-commerce, businesses increasingly rely on data-driven customer segmentation to personalize experiences and optimize marketing.

• Traditional RFM models (Recency, Frequency, Monetary) are commonly used but limited to short-term behavior and overlook long-term engagement.

II. Problem with Traditional Models

• RFM fails to capture:

  • Customer longevity

  • Cumulative spending

  • Evolving behavioral patterns

• The LRFS model (Length, Recency, Frequency, Spend) improved on this by adding:

  • Length – duration of customer relationship

  • Spend – total spend over time

• However, LRFS still lacks per-transaction insights—i.e., Monetary value per transaction.

III. Proposed Solution: LRFSM Model

• The study proposes LRFSM: Length, Recency, Frequency, Spend, and Monetary value.

• This model captures both cumulative and per-transaction behavior, allowing businesses to:

  • Distinguish between frequent low spenders and infrequent high spenders

  • Perform more precise customer segmentation

IV. Methodology

• Synthetic dataset simulating realistic e-commerce transactions over two years.

• Features engineered:

  • L (Length) = Last Txn – First Txn + 1

  • R (Recency) = Reference Date – Last Txn

  • F (Frequency) = Number of transactions

  • S (Spend) = Total spend

  • M (Monetary) = Spend / Frequency

• Normalization using Z-score to standardize features.

• Dimensionality reduction via PCA for visualization and computational efficiency.

V. Clustering Algorithms Used

The study evaluated multiple unsupervised ML algorithms:

1. KMeans & MiniBatch KMeans – Scalable, centroid-based

2. Agglomerative Clustering – Hierarchical approach

3. Gaussian Mixture Models (GMM) – Probabilistic clusters

4. DBSCAN – Density-based, identifies outliers

5. Spectral Clustering – Graph-based, handles complex cluster shapes

6. KShape – Suitable for time-series behavior

VI. Evaluation Metrics

Unsupervised Metrics:

• Silhouette Score – Measures cluster cohesion/separation

• Davies-Bouldin Index (DBI) – Lower is better

• Calinski-Harabasz Index (CHI) – Higher is better

Supervised Metrics (when labels available):

• Accuracy, Precision, Recall, F1-Score

• Hungarian algorithm used for label alignment

VII. Results & Insights

• DBSCAN: Identified dense central clusters and isolated outliers, but sensitive to parameter tuning.

• Spectral Clustering: Produced well-separated, balanced clusters, suitable for complex data.

• KShape: Generated compact, low-overlap clusters—effective for distinguishing temporal patterns, especially high-spend users.

Cluster Profiling was done by analyzing average LRFSM feature values per cluster, enabling:

• Identification of high-value loyal customers

• Discovery of new or at-risk customers

• Support for personalized promotions and churn prevention

VIII. Related Work

• Builds on LRFS by Ameen Al-Dubai et al. (KMeans-based segmentation).

• Other research introduced:

  • Affinity-based spectral clustering

  • Regularized KMeans to handle high-dimensional data

  • CLV-based models for long-term value segmentation

  • Deep Embedded Clustering (DEC) for mixed-data clustering

  • Hybrid churn prediction models combining statistical and ML techniques

IX. Future Directions

• Integration of deep learning clustering (e.g., DEC, X-DEC) for better handling of:

  • Mixed data types

  • High-dimensional features

• Focus on real-world deployment for adaptive, automated customer engagement systems.

References

[1] Al-Dubai, T. Al-Khalifa, S. Ahsan, R. F. Olanrewaju, and D. Yousif, “LRFS: Online Shoppers Behavior-Based Efficient Customer Segmentation Model,” International Journal of Advanced Computer Science and Applications (IJACSA), vol. 14, no. 4, pp. 579–586, 2023. [2] T. Lin, L. Sun, C. Guo, and Z. Zhang, “Adaptive Spectral Affinity Propagation Clustering,” Proceedings of the 2021 International Conference on Artificial Intelligence and Computer Engineering (ICAICE), 2021, pp. 478–483. [3] H.-H. Zhao, X.-C. Luo, R. Ma, and X. Lu, “An Extended Regularized K-Means Clustering Approach for High-Dimensional Customer Segmentation With Correlated Variables,” IEEE Access, vol. 9, pp. 112508–112520, 2021. [4] M. B. Hosseini and M. J. Tarokh, “Customer Segmentation Using CLV Elements,” International Journal of Engineering Research and Applications (IJERA), vol. 6, no. 2, pp. 68–74, 2016. [5] J. Liao, A. Jantan, Y. Ruan, and C. Zhou, “Multi-Behavior RFM Model Based on Improved OM Neural Network Algorithm for Customer Segmentation,” IEEE Access, vol. 9, pp. 169389–169400, 2021. [6] A. Naeem, A. Shabbir, N. U. Hassan, C. Yuen, A. Ahmad, and W. Tushar, “Understanding Customer Behavior in Multi-Tier Demand Response Management Program,” IEEE Access, vol. 8, pp. 170717–170733, 2020. [7] J. W. T. M. de Kok et al., \"Deep embedded clustering generalisability and adaptation for integrating mixed datatypes: two critical care cohorts,\" Scientific Reports, vol. 14, no. 1045, 2024, doi: 10.1038/s41598-024-51699-z. [8] Y. G. Lee, C. W. Park, and S. J. Kang, \"Deep Embedded Clustering Framework for Mixed Data,\" IEEE Access, vol. 10, pp. 1–12, 2022, doi: 10.1109/ACCESS.2022.3232372. [9] X. D. Wang, R. C. Chen, F. Yan, Z. Q. Zeng, and C. Q. Hong, \"Fast Adaptive K-Means Subspace Clustering for High-Dimensional Data,\" IEEE Access, vol. 7, pp. 42639–42652, 2019, doi: 10.1109/ACCESS.2019.2907043. [10] N. T. Lee, H. C. Lee, J. Hsin, and S. H. Fang, \"Prediction of Customer Behavior Changing via a Hybrid Approach,\" IEEE Open Journal of the Computer Society, vol. 1, pp. 1–12, 2023, doi: 10.1109/OJCS.2023.3336904. [11] W. M. Wong and W. Su, \"Segmenting Online Shoppers: A Combined Cluster and Logistic Regression Approach for Forecasting Purchase Behavior,\" IEEE Access, vol. XX, pp. 1–10, 2025, doi: 10.1109/ACCESS.2025.3565897.

Copyright

Copyright © 2025 Sravan Kumar Mandeti. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.