Educational Resource Recommendation Model Based on Compressed Interaction and Feature Weighting

An Haonan

The intelligent education industry is an important branch of China’s education and an important part of information construction. However, in the implementation of the learner-centered large-scale online education platform in smart education, there are still problems such as “information trek” and “information overload” brought by massive and abundant educational resources for learners. To solve these problems, the Compressed Interaction and Feature Weight Network (CIFW) is proposed as the core of personalized recommendation technology. The framework structure of the model adopts parallel mode. After the input features are entered into the compressed interaction module, the feature weighting module and the linear network respectively, the operation of high-order feature interaction and the operation of assigning different weights to the features enables the learning of more nonlinear relations. Moreover, the generalization ability and memory ability of the output results are enhanced by combining the results generated by nonlinear and linear. Compared with the original XDeepFM model, the ACC value of this model is increased by 0.00605 and 0.00213, and the AUC value is increased by 0.00678 and 0.00388, respectively, to achieve a more efficient recommendation effect.

Select Volume / Issues:

• Volume 14, Issue 5, Sept. 2023

Year:

2023

Type of Publication:

Article

Keywords:

Deep Learning; Intelligent Education; Recommendation Algorithm

Journal:

IJECCE

Volume:

14

Number:

5

Pages:

36-43

Month:

September

ISSN:

2249-071X