Scholarlink

<p>This paper investigates the effectiveness of classical association rule mining methods (Apriori and FP-Growth), an autoencoder-based approach (ARM-AE), and a deep CTR model (DCNv3) for personalized SMS marketing. While Apriori and FP-Growth identify frequent item associations with high support and confidence, their exhaustive nature often leads to redundant rules that are less suitable for targeted personalization. DCNv3 achieves strong engagement performance through end-to-end prediction but lacks interpretability. Experimental results show that ARM-AE provides the best trade-off between engagement and explainability by extracting a compact set of high-confidence and reliable association rules. The approach is evaluated on a proprietary dataset as well as on public benchmark datasets, Dunnhumby and Avazu, demonstrating consistent performance and strong generalization under sparse and high-dimensional conditions. Based on these results, ARM-AE is selected as the knowledge extraction component and integrated into a retrieval-augmented generation (RAG) framework to guide large language models in generating personalized, context-aware, and action-oriented SMS messages. Overall, the proposed ARM-AE + RAG framework offers an effective and interpretable solution for data-driven and scalable personalized marketing.</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 product recommendation using market basket patterns and LLM-powered retrieval systems

  • 期刊论文

  • 发表日期：

    2026年4月2日

查看全文
  • Rania Bengharsallah,
  • Mounira Tarhouni,
  • Faten Ben Aicha

摘要

This paper investigates the effectiveness of classical association rule mining methods (Apriori and FP-Growth), an autoencoder-based approach (ARM-AE), and a deep CTR model (DCNv3) for personalized SMS marketing. While Apriori and FP-Growth identify frequent item associations with high support and confidence, their exhaustive nature often leads to redundant rules that are less suitable for targeted personalization. DCNv3 achieves strong engagement performance through end-to-end prediction but lacks interpretability. Experimental results show that ARM-AE provides the best trade-off between engagement and explainability by extracting a compact set of high-confidence and reliable association rules. The approach is evaluated on a proprietary dataset as well as on public benchmark datasets, Dunnhumby and Avazu, demonstrating consistent performance and strong generalization under sparse and high-dimensional conditions. Based on these results, ARM-AE is selected as the knowledge extraction component and integrated into a retrieval-augmented generation (RAG) framework to guide large language models in generating personalized, context-aware, and action-oriented SMS messages. Overall, the proposed ARM-AE + RAG framework offers an effective and interpretable solution for data-driven and scalable personalized marketing.