Lv Sheng, Wang Jiabin, Deng Fan, Yan Penggui
school of engineering, Huaqiao University, 362011, Quanzhou, Fujian, China.
Sci Rep. 2024 Jul 25;14(1):17119. doi: 10.1038/s41598-024-66393-3.
In the realm of e-commerce, personalized recommendations are a crucial component in enhancing user experience and optimizing sales efficiency. To address the inherent sparsity challenge prevalent in collaborative filtering algorithms within personalized recommendation systems, we propose a novel hybrid e-commerce recommendation algorithm based on the User-Nearest-Neighbor model. By integrating the user nearest neighbor model with other recommendation algorithms, this approach effectively mitigates data sparsity and facilitates a more nuanced understanding of the user-product relationship, consequently elevating recommendation quality and enhancing user experience. Taking into account considerations such as data scale and recommendation performance, we conducted experiments utilizing the Spark distributed platform. Empirical findings demonstrate the superiority of our hybrid algorithm over standalone collaborative filtering algorithms across various recommendation indicators.
在电子商务领域，个性化推荐是提升用户体验和优化销售效率的关键组成部分。为了解决个性化推荐系统中协同过滤算法普遍存在的固有稀疏性挑战，我们提出了一种基于用户最近邻模型的新型混合电子商务推荐算法。通过将用户最近邻模型与其他推荐算法相结合，这种方法有效地缓解了数据稀疏性，并有助于更细致地理解用户与产品的关系，从而提高推荐质量并提升用户体验。考虑到数据规模和推荐性能等因素，我们利用Spark分布式平台进行了实验。实证结果表明，在各种推荐指标上，我们的混合算法优于独立的协同过滤算法。
