Scholarlink

This paper proposes DeepHUPM, a novel deep learning-based framework for high-utility pattern mining (HUPM) in large-scale datasets. By integrating deep neural networks (DNNs), LSTM models with attention mechanisms, and distributed processing via Apache Spark, DeepHUPM effectively captures temporal dependencies and optimizes mining performance. The proposed model achieves up to 98.5% accuracy, 40–50% reduced memory usage, and improved runtime efficiency compared to traditional methods such as EFIM and DMOUM. Unlike existing approaches, DeepHUPM incorporates an interpretable attention layer and demonstrates superior performance over recent Transformer-based baselines. Additionally, the framework’s scalability is evaluated across varying dataset sizes and Spark cluster configurations, confirming its robustness in real-world big data environments.

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

Deep Learning Based Model Optimization Method for Mining High Utility Patterns in Large Datasets

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Arkan A. Ghaib,
  • Amna Kadhim Ali,
  • Jawad Alkenani,
  • Roaa M. Salih

摘要

This paper proposes DeepHUPM, a novel deep learning-based framework for high-utility pattern mining (HUPM) in large-scale datasets. By integrating deep neural networks (DNNs), LSTM models with attention mechanisms, and distributed processing via Apache Spark, DeepHUPM effectively captures temporal dependencies and optimizes mining performance. The proposed model achieves up to 98.5% accuracy, 40–50% reduced memory usage, and improved runtime efficiency compared to traditional methods such as EFIM and DMOUM. Unlike existing approaches, DeepHUPM incorporates an interpretable attention layer and demonstrates superior performance over recent Transformer-based baselines. Additionally, the framework’s scalability is evaluated across varying dataset sizes and Spark cluster configurations, confirming its robustness in real-world big data environments.