Deep reinforcement learning-based resource recommendation system for ideological and political education
- 1Department of Law, Semyung University
- 2Enterprise Compliance Research Center, Binzhou Polytechnic
Abstract
In the context of developments in digital education and artificial intelligence technologies, ideological and political education (IPE) scenarios (e.g., university teaching, enterprise training, and community practice) generate growing demands for accurate resource recommendations. Nevertheless, most existing methods rely on static algorithms; they fail to depict nonlinear relationships between user behaviors and resource features and cannot satisfy the personalized learning demands of multiple scenarios. Therefore, this study proposes an IPE recommendation system based on the Soft Actor-Critic (SAC) (SAC is a deep reinforcement learning (DRL) algorithm). This system integrates behavioral data (such as user browsing, learning duration, and ratings) with resource features. It dynamically optimizes recommendation strategies via a state-action-reward mechanism and introduces a discount factor to balance immediate and long-term benefits for intelligent resource matching. Experiments are conducted for verification based on three types of IPE datasets from universities, enterprises, and communities. The results show that on the university dataset, the algorithm achieves Hit Ratio at 10 (HR@10) of 0.6025 and Normalized Discounted Cumulative Gain at 10 (NDCG@10) of 0.381; on the enterprise dataset, corresponding indicators reach 0.47 and 0.3715; on the community dataset, HR@10 and NDCG@10 are 0.5942 and 0.376. All indicator values outperform comparative methods, including Linear Upper Confidence Bound (LinUCB), Deep Q-Network (DQN), Dynamic Graph-based Recommendation (DGRec), Knowledge-aware Collaborative Graph Network (KCGN), and Multi-channel Hypergraph Convolutional Network (MHCN). The model achieves stable convergence after around 400 training episodes, and the average reward is 0.390; meanwhile, it takes the lead on long-term indicators such as the 30-day user repeated learning rate. These findings demonstrate that the SAC-based DRL method effectively improves the accuracy, stability, and dynamic adaptability of IPE resource recommendations; it also provides theoretical foundations and practical references for the digital development of multi-scenario IPE.
Related
- Federated graph learning with spatio-temporal dynamics for cross-border recommendation
- Reinforcement learning–coupled neural ode modelling reveals ferroptotic resolution failure as a shared dynamical axis linking periodontitis and coronary atherogenesis
- DEEP REINFORCEMENT LEARNING IN ONCOLOGY: ADVANCES IN CANCER IMAGING, RADIOTHERAPY, AND PERSONALIZED TREATMENT
- Reinforcement learning as a foundation for psychodynamics
- Multiproperty optimization of therapeutic antibodies through sequence editing via large language model and reinforcement learning
BibTeX
@article{song2026deep,
title = {{Deep reinforcement learning-based resource recommendation system for ideological and political education}},
author = {Zixuan Song and Lianxin Geng and Xiao Wang and Xiaoke Gong},
journal = {Scientific Reports},
year = {2026},
month = jul,
doi = {10.1038/s41598-026-63860-x},
url = {https://doi.org/10.1038/s41598-026-63860-x},
note = {CC BY},
}
RIS
TY - JOUR
TI - Deep reinforcement learning-based resource recommendation system for ideological and political education
AU - Zixuan Song
AU - Lianxin Geng
AU - Xiao Wang
AU - Xiaoke Gong
PY - 2026
DA - 2026/07/01
JO - Scientific Reports
DO - 10.1038/s41598-026-63860-x
UR - https://doi.org/10.1038/s41598-026-63860-x
AB - In the context of developments in digital education and artificial intelligence technologies, ideological and political education (IPE) scenarios (e.g., university teaching, enterprise training, and community practice) generate growing demands for accurate resource recommendations. Nevertheless, most existing methods rely on static algorithms; they fail to depict nonlinear relationships between user behaviors and resource features and cannot satisfy the personalized learning demands of multiple scenarios. Therefore, this study proposes an IPE recommendation system based on the Soft Actor-Critic (SAC) (SAC is a deep reinforcement learning (DRL) algorithm). This system integrates behavioral data (such as user browsing, learning duration, and ratings) with resource features. It dynamically optimizes recommendation strategies via a state-action-reward mechanism and introduces a discount factor to balance immediate and long-term benefits for intelligent resource matching. Experiments are conducted for verification based on three types of IPE datasets from universities, enterprises, and communities. The results show that on the university dataset, the algorithm achieves Hit Ratio at 10 (HR@10) of 0.6025 and Normalized Discounted Cumulative Gain at 10 (NDCG@10) of 0.381; on the enterprise dataset, corresponding indicators reach 0.47 and 0.3715; on the community dataset, HR@10 and NDCG@10 are 0.5942 and 0.376. All indicator values outperform comparative methods, including Linear Upper Confidence Bound (LinUCB), Deep Q-Network (DQN), Dynamic Graph-based Recommendation (DGRec), Knowledge-aware Collaborative Graph Network (KCGN), and Multi-channel Hypergraph Convolutional Network (MHCN). The model achieves stable convergence after around 400 training episodes, and the average reward is 0.390; meanwhile, it takes the lead on long-term indicators such as the 30-day user repeated learning rate. These findings demonstrate that the SAC-based DRL method effectively improves the accuracy, stability, and dynamic adaptability of IPE resource recommendations; it also provides theoretical foundations and practical references for the digital development of multi-scenario IPE.
SN - 2045-2322
ER -