基于深度学习的电子商务个性化推荐模型

Personalized Recommendation Model of E-Commerce Based on Deep Learning

智能推荐是电子商务领域的一项重要任务。目前普遍使用关联规则、协同过滤、马尔可夫链、递归神经网络等技术进行购物篮推荐。本文主要研究基于深度学习的电子商务智能推荐系统(IRS)。本文首先进行了电子商务推荐系统的总体设计,提出了电子商务IRS的功能模块和系统架构。然后,讨论了电子商务IRS中的推荐算法,并基于卷积神经网络对电子商务IRS进行了优化。最后,本文比较分析了三种流行的推荐算法在阿里巴巴数据集上的性能。实验结果表明,本模型在不同推荐列表长度下的召回率和NDCG指标上均取得了更高的数值,均明显优于其他两种算法(Item和BPR),在挖掘用户会话序列中的兴趣和行为偏好方面具有很强的价值,能够从大规模数据中学习到更丰富和复杂的用户行为和商品信息,这有助于提高推荐的效率和准确性,具有较强的实际意义和推广价值。

Intelligent recommendation is an important task in the field of electronic commerce. At present, association rules, collaborative filtering, Markov chain and recursive neural network are widely used to recommend shopping baskets. This paper mainly studies the e-commerce intelligent recommendation system (IRS) based on deep learning. In this paper, firstly, the overall design of e-commerce recommendation system is carried out, and the functional modules and system architecture of e-commerce IRS are put forward. Then, the recommendation algorithm in e-commerce IRS is discussed, and the e-commerce IRS is optimized based on convolutional neural network. Finally, this paper compares and analyzes the performance of three popular recommendation algorithms on Alibaba data sets. The experimental results show that this model achieves higher recall and NDCG index under different recommended list lengths, which are obviously superior to the other two algorithms (Item and BPR). It is of great value in mining users’ interests and behavior preferences in conversation sequences, and can learn richer and more complex information about users’ behaviors and commodities from large-scale data, which is helpful to improve the efficiency and accuracy of recommendation, and has strong practical significance and promotion value.