Scholarlink

<p>Sequential patent recommendation for alliance firms must account for technological diversity and peer influence, yet existing methods overlook these aspects. Technological diversity is essential because firms typically pursue multiple technological directions to enhance their adaptability and competitive edge. Additionally, in strategic alliances, peer influence plays a crucial role as firms are influenced by the technological advancements of their partners, shaping their own patent acquisition decisions. To bridge the gap, we propose a deep learning-based approach called technological Diversity and peer Influence-aware Patent Recommendation (DIPR). DIPR integrates a capsule network for technological diversity and a graph convolutional network for peer influence. Evaluated on real-world datasets from biotechnology (156 companies, 10,034 patents) and medical equipment alliances (131 companies, 7,237 patents), DIPR achieves significant improvements: modeling technological diversity enhances precision, recall, F1, and NDCG by 2.98%, 2.88%, 2.88%, and 1.60%, respectively, while incorporating peer influence boosts these metrics by 10.47%, 9.19%, 9.46%, and 8.46%. DIPR outperforms state-of-the-art baselines by 10.22%–15.05% across metrics, demonstrating its efficacy in aligning recommendations with firms’ evolving needs and alliance dynamics.</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

Sequential patent recommendation for alliance firms: a deep learning approach integrating technological diversity and peer influence

  • 期刊论文

  • 发表日期：

    2026年4月30日

查看全文
  • Zhaobin Liu,
  • Weiwei Deng,
  • Jicheng Zeng,
  • WeiJing Zhu,
  • Jian Ma

摘要

Sequential patent recommendation for alliance firms must account for technological diversity and peer influence, yet existing methods overlook these aspects. Technological diversity is essential because firms typically pursue multiple technological directions to enhance their adaptability and competitive edge. Additionally, in strategic alliances, peer influence plays a crucial role as firms are influenced by the technological advancements of their partners, shaping their own patent acquisition decisions. To bridge the gap, we propose a deep learning-based approach called technological Diversity and peer Influence-aware Patent Recommendation (DIPR). DIPR integrates a capsule network for technological diversity and a graph convolutional network for peer influence. Evaluated on real-world datasets from biotechnology (156 companies, 10,034 patents) and medical equipment alliances (131 companies, 7,237 patents), DIPR achieves significant improvements: modeling technological diversity enhances precision, recall, F1, and NDCG by 2.98%, 2.88%, 2.88%, and 1.60%, respectively, while incorporating peer influence boosts these metrics by 10.47%, 9.19%, 9.46%, and 8.46%. DIPR outperforms state-of-the-art baselines by 10.22%–15.05% across metrics, demonstrating its efficacy in aligning recommendations with firms’ evolving needs and alliance dynamics.