Scholarlink

In today's digital environment, marketplaces and online services generate large amounts of customer data on a regular basis. Analysing the information is essential to comprehending customer behaviour and making sensible choices. Customer segmentation is a crucial method for grouping customers based on similar features.However traditional clustering algorithms like K-means sometimes fails when tackling with complex and non-convex structured data.Although Spectral clustering is superior in such situations, thes result is dependent on the appropriate selection of the RBF kernel scale parameter that is frequently obtained through manual tuning.To address this drawbacks, the research proposes a Deep Q-Networks(DQN) based spectral clustering technique for automated optimization of parameter with the goal to get over the limitations. By optimizing the silhouette score, the DQN agent develops to rapidly change in the parameter tuning process that is presented as a reinforcement learning problem.The proposed approach delivers more accurate and significant customer groupings than traditional K-means clustering.

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

DQN-Spectral: Reinforcement Learning Guided Spectral Clustering for Customer Segmentation for Marketing

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • S. Devatharshini,
  • K. Mouthami,
  • Sharli Iris John Baskar,
  • N. Subasri

摘要

In today's digital environment, marketplaces and online services generate large amounts of customer data on a regular basis. Analysing the information is essential to comprehending customer behaviour and making sensible choices. Customer segmentation is a crucial method for grouping customers based on similar features.However traditional clustering algorithms like K-means sometimes fails when tackling with complex and non-convex structured data.Although Spectral clustering is superior in such situations, thes result is dependent on the appropriate selection of the RBF kernel scale parameter that is frequently obtained through manual tuning.To address this drawbacks, the research proposes a Deep Q-Networks(DQN) based spectral clustering technique for automated optimization of parameter with the goal to get over the limitations. By optimizing the silhouette score, the DQN agent develops to rapidly change in the parameter tuning process that is presented as a reinforcement learning problem.The proposed approach delivers more accurate and significant customer groupings than traditional K-means clustering.