Abstract: To address the problem of recommendation algorithm computing readers similarity or books similarity with low accuracy and recommendation quality, Personalized recommendation algorithm based on multi-feature is proposed. It computes books similarity based on Chinese library classification method and books feature vector, computes readers similarity based on readers feature vector and borrow records. Based on this, two prediction results are produced. The last recommendation is produced to readers by weight of this two prediction results. Experimental result shows that the proposed algorithm achieves more recommendation accuracy on books.

Key words: Chinese library classification method, book feature vector, reader feature vector, similarity, recommendation algorithm, Chinese library classification tree, professional classification tree

摘要： 现有推荐算法计算读者之间或图书之间的相似性不准确、推荐精确度不高。为此，提出一种基于多特征的个性化图书推荐算法。根据中图分类法及图书的特征向量计算图书的相似性，依据读者的特征向量及借阅记录计算读者的相似性。在此基础上产生2种预测结果并对其进行加权，产生最终推荐。实验结果表明，该算法具有较高的图书推荐精确度。

关键词: 中图分类法, 图书特征向量, 读者特征向量, 相似性, 推荐算法, 中图分类树, 专业分类树

CLC Number: 

• TP301.6