Personalized learning path optimization based on enhanced deep neural network: higher education teaching model integrating learner behavior and cognitive style
Abstract As information technology and artificial intelligence advance rapidly, personalized learning has emerged as a key approach to enhancing teaching quality and learning outcomes in higher education. Current research considers behavior or cognitive style separately, lacks effective integration,...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-08-01
|
| Series: | Discover Artificial Intelligence |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s44163-025-00467-7 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract As information technology and artificial intelligence advance rapidly, personalized learning has emerged as a key approach to enhancing teaching quality and learning outcomes in higher education. Current research considers behavior or cognitive style separately, lacks effective integration, and traditional methods are mostly based on static analysis, which makes it difficult to cope with students' dynamic needs and real-time feedback, and lacks flexibility and adaptability. To address the above problems, this paper designs an effective model for optimizing personalized learning paths using Enhanced Deep Neural Networks (EDNN). The model adopts the Actor-Critic framework to introduce a multilayer perceptron (MLP) in the Actor part to fuse behavioral data and cognitive style, and a long short-term memory (LSTM) neural network in the Critic part to process time series data, thereby realizing the selection and dynamic adjustment of personalized learning paths. Experiments show that the model proposed in this study can integrate learners' behavioral data and cognitive styles to optimize personalized learning paths. Compared with the Q-learning model and the Deep Q-learning Network (DQN) model, the model in this study is superior in learning speed, stability, adaptability, etc. During the experiment, the model studied in this paper can reduce the loss to 0.1 after 40 iterations and converge quickly in the later stage. The shortest average feedback response time of 100 experiments is 1.123 s, and the shortest average path adjustment calculation time is 2.010 s. The recommendation effect of different deep neural networks (DNN) on personalized learning paths is analyzed. This model outperforms others in terms of accuracy for personalized learning path recommendations. The model can efficiently integrate learners' behavioral data and cognitive styles, adjust learning paths in real-time, and demonstrate rapid response and optimization capabilities in practical applications. |
|---|---|
| ISSN: | 2731-0809 |