【摘要】 由于数字图书馆中的书目数量通常非常庞大，且书目个性化推荐过程易受冗余信息、书目类别、不同文本信息等问题的影响，导致书目推荐质量差。为此，文章提出考虑冗余干扰的数字图书馆书目个性化推荐算法。采用粗糙集实现书目数据清洗。应用词频-逆文本频率指数（Term Frequency—Inverse Document Frequency,TF-IDF）算法提取书目文本特征，将自组织映射神经网络（Self-Organizing Map,SOM）网络模型与K-means聚类算法相结合，通过联合聚类缩小待预测书目资源的数量与搜索范围，提供符合读者偏好的书目，完成数字图书馆书目个性化的推荐。实验结果显示，所提算法的推荐质量高，平均绝对误差（Mean Absolute Error,MAE）仅为0.3，文本特征提取概率可达0.7以上，推荐的书目类别更具多样性，推荐性能更优。更多还原

【Abstract】 Due to the typically vast amount of bibliography in digital libraries, and the susceptibility of personalized recommendation process to redundant information, bibliographic categories and different text information, the quality of bibliographic recommendation often suffers. Therefore, a personalized recommendation algorithm considering redundancy interference for digital library bibliography is proposed. Rough sets are employed to clean the bibliographical data. The TF-IDF(Term Frequence-Inverse Document Frequency) algorithm is used to extract bibliographic text features, and SOM(Self-Organizing Map) network model and K-means clustering algorithm are combined to narrow the number and search scope of the bibliographic resources to be predicted through joint clustering, thereby providing books that meet readers’ preference and completing personalized bibliographic recommendation in digital libraries. It shows that the the proposed algorithm has achived high recommendation quality, with MAE(Mean Absolute Error) of only 0.3, probability of text feature extraction exceeding 0.7, greater diversity in recommended bibliographic categories, and superior overall recommendation performance.更多还原