Research Article
A Deep Reinforcement Learning-Based Approach for Intelligent Recommendation of Digital Museums
@INPROCEEDINGS{10.1007/978-3-032-00303-4_14, author={Fang Liu and Fei Dai and Lilian Lee Shiau Gee and Imelda Ann Achin and Na Zhang}, title={A Deep Reinforcement Learning-Based Approach for Intelligent Recommendation of Digital Museums}, proceedings={Advanced Hybrid Information Processing. 8th International Conference, ADHIP 2024, Jiaxing, China, September 20--22, 2024, Proceedings, Part III}, proceedings_a={ADHIP}, year={2026}, month={11}, keywords={Deep reinforcement learning Intangible Cultural Heritage Digital museum ItemCF Intelligent recommendation}, doi={10.1007/978-3-032-00303-4_14} }- Fang Liu
Fei Dai
Lilian Lee Shiau Gee
Imelda Ann Achin
Na Zhang
Year: 2026
A Deep Reinforcement Learning-Based Approach for Intelligent Recommendation of Digital Museums
ADHIP
Springer
DOI: 10.1007/978-3-032-00303-4_14
Abstract
As an important platform for cultural dissemination, digital museums have intelligent recommendation systems that can more effectively disseminate cultural knowledge. In order to improve recommendation effectiveness and user satisfaction, a digital museum intelligent recommendation method based on deep reinforcement learning is proposed. Acquire user preferences based on mixed evaluation methods. Collect exhibition information of intangible cultural heritage digital museum and design exhibition information database for query and analysis. On this basis, in-depth reinforcement learning is used to represent and learn user behavior and exhibit information, and useful features are extracted from user behavior and exhibit information. Based on Item Collaborative Filtering (ItemCF), an intelligent recommendation algorithm for intangible cultural heritage digital museums is designed. Use Mahout open source data mining tool library to design an intelligent recommendation engine module for intangible cultural heritage digital museums. The test results show that this method has better frontiers in accuracy, recall, and F1 score compared to traditional methods, and can more accurately capture users’ interests and preferences, and generate more personalized recommendation lists for users.