ANN方法在网络推荐算法中的应用

Application of the ANN Method in the Network Recommendation Algorithm

摘要: 文章研究探讨了人工神经网络(ANN)在网络推荐算法领域的运用。ANN算法具备将用户和物品的特征向量映射至低维空间的能力，通过衡量用户过往行为与潜在推荐对象间的相似程度，迅速锁定用户可能青睐的内容，进而实现个性化推荐。研究过程中，我们运用ANN方法对用户的浏览、收藏及加购行为的权值展开了动态解析与计算。重点关注于多层人工神经网络模型，购买模型和基于距离的向量匹配算法相结合，并在模型中提出高效且准确的个性化推荐系统。提出感知器模型，进而提出多层人工神经网络，进行网络预测，不仅解决了信息茧房与过度推荐的问题，还解决了用户反馈难以进行个性化推荐的问题，并以此提出一种基于距离的向量匹配算法，针对不同用户推送个性化的商品。实验阶段，我们依据用户对商品的收藏频次、购买相关商品的次数以及浏览该商品及其相关商品的数量，来预测用户即将购买的商品数量及其在各商品上的浏览时长。将网络预测的结果与实际测试集数据进行比对后，我们发现网络预测与实际情况展现出较高的吻合度。随后采用基于距离的向量匹配技术，针对不同用户推送个性化的商品信息。最后进行对比分析，与其他网络预测方法比较，突出本文方法的优势。

Abstract: In this paper, We studies the application of ANN (artificial neural network) method in network recommendation algorithm. ANN algorithm can map the feature vectors of users and objects into the low dimensional space, by calculating the similarity between user historical behavior and candidate recommendation items, quickly find the content that the user may be interested in, and realize personalized recommendation.In the process of research, we used the ANN method to dynamically analyse and calculate the weights of users’ browsing, collection and purchase behaviours. This paper focuses on the combination of multi-layer artificial neural network model, purchase model and distance-based vector matching algorithm, and proposes an efficient and accurate personalised recommendation system in the model. The perceptron model is proposed, and then a multi-layer artificial neural network is proposed for network prediction, which not only solves the problem of information cocoon and over-recommendation, but also solves the problem that it is difficult to make personalised recommendations based on user feedback, and proposes a distance-based vector matching algorithm to push personalised products for different users. At the same time, making personalized recommendations using ANN methods can gain new cognition, create new wisdom, and produce more valuable decisions. Explore the consumer needs of the users from the historical data, explore the new cognition from the old data, create the new wisdom from the new cognition combined with the machine learning algorithms, and finally help the users to find the goods they are interested in, and present the most suitable products to the users. Finally, a comparative analysis is conducted to highlight the advantages of our method compared to other network prediction methods.