We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.IR

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Information Retrieval

Title: URIR: Recommendation algorithm of user RNN encoder and item encoder based on knowledge graph

Abstract: Due to a large amount of information, it is difficult for users to find what they are interested in among the many choices. In order to improve users' experience, recommendation systems have been widely used in music recommendations, movie recommendations, online shopping, and other scenarios. Recently, Knowledge Graph (KG) has been proven to be an effective tool to improve the performance of recommendation systems. However, a huge challenge in applying knowledge graphs for recommendation is how to use knowledge graphs to obtain better user codes and item codes. In response to this problem, this research proposes a user Recurrent Neural Network (RNN) encoder and item encoder recommendation algorithm based on Knowledge Graph (URIR). This study encodes items by capturing high-level neighbor information to generate items' representation vectors and applies an RNN and items' representation vectors to encode users to generate users' representation vectors, and then perform inner product operation on users' representation vectors and items' representation vectors to get probabilities of users interaction with items. Numerical experiments on three real-world datasets demonstrate that URIR is superior performance to state-of-the-art algorithms in indicators such as AUC, Precision, Recall, and MRR. This implies that URIR can effectively use knowledge graph to obtain better user codes and item codes, thereby obtaining better recommendation results.
Subjects: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
Cite as: arXiv:2111.00739 [cs.IR]
  (or arXiv:2111.00739v1 [cs.IR] for this version)

Submission history

From: Zhen Long [view email]
[v1] Mon, 1 Nov 2021 07:28:11 GMT (86kb,D)

Link back to: arXiv, form interface, contact.