Traditional course recommendation methods lack flexibility and perform poorly, as their quality relies heavily on historical datasets -making them ill-suited for complex-attribute tasks.This study integrates collaborative filtering (CF) with deep learning, leveraging item-based recommendation timeliness to boost accuracy.Experimental outcomes indicate that the proposed algorithm achieves a mean square error of 0.572, whereas the mean square error for content-based recommendation algorithms and singular value decomposition methods increase by 0.076 and 0.099.The core new insight lies in the bidirectional empowerment fusion architecture of CF and deep learning (DL): integrating CF-derived course similarity as a priori information into the DL model's input and attention mechanism, which not only solves CF's sensitivity to sparse data but also compensates for DL's poor interpretability and reliance on massive data.This provides a novel hybrid recommendation paradigm for addressing personalised course recommendation issues in higher vocational colleges.
Xiaoguang Liu (Thu,) studied this question.