【摘要】 合理的技术创新合作预测方法是企业寻找合适的技术创新合作伙伴以提升技术创新绩效的有效手段。论文基于企业专利数据，构建专利所有权人共现网络，应用Katz指标计算企业之间的路径相似性，应用TF-IDF算法构建企业关键词向量，结合余弦相似性计算企业之间的内容相似性，应用社会网络分析方法中的中心性指标计算企业的位置相似性，将三者进一步融合得到企业之间合作的潜在可能性。通过对石墨烯领域企业专利数据分析预测企业间合作的可能性，证实该方法有效，AUC指标值为0.7242,优于单一指标相似性推荐方法，能够提升合作推荐中合适匹配的精确度。更多还原

【Abstract】 The method of predicting reasonable technology innovation cooperation is an effective approach for enterprises to identify suitable partners in technology innovation, thereby enhancing their performance in this area.The present study utilizes enterprise patent data to construct a co-occurrence network of patent owners.It applies the Katz index to calculate the path similarity between enterprises, employs the TF-IDF algorithm to construct an enterprise keyword vector, calculates content similarity between enterprises using cosine similarity, and utilizes centrality indices from social network analysis methods to determine location similarity among enterprises.The enhanced integration of the three entities will unlock the collaborative potential among enterprises.Through the analysis of enterprise patent data in the Graphene field, we predict potential collaborations between enterprises and demonstrate the effectiveness of this method.The AUC index value is 0.7242,surpassing that of a single-index similarity recommendation method, thereby enhancing the accuracy of suitable matches in collaboration recommendations.更多还原