融合协同过滤的XGBoost在音乐推送上的应用研究

该文研究一种基于融合协同过滤和XGBoost的音乐推荐算法。首先,使用协同过滤算法计算用户之间或物品之间的相似度,从而得到初始的推荐列表,作为召回集。考虑到协同过滤法产生的音乐推荐列表还存在计算量大、稀疏性等问题,导致推荐列表并没有那么准确。接下来,对推荐列表中的每个项目进行特征提取和特征工程,并使用XGBoost算法对其进行预测,得到最终的推荐列表。该研究的贡献在于提出一种新的音乐推送算法,融合协同过滤和XGBoost算法的优点,可以得到更精准的音乐推荐列表。

This paper studies a music recommendation algorithm based on fusion collaborative filtering and XGBoost.First,we used a collaborative filtering algorithm to calculate the similarity between users or items and obtained an initial list of recommendations as a recall set.Considering that the music recommendation list generated by collaborative filtering method still has some problems such as large computation and sparsity,the recommendation list is not so accurate.Next,we carried out feature extraction and feature engineering for each item in the recommendation list,and used XGBoost algorithm to predict it and got the final recommendation list.The contribution of this study is to propose a new music recommendation algorithm,which combines the advantages of collaborative filtering and XGBoost algorithm to get more accurate music recommendation list.