Abstract: The product service system is a personalized solution provided by the enterprise for the user.To design the solution quickly and accurately,an improved collaborative filtering algorithm was proposed to recommend the solution according to the user's personalized requirements.In the traditional collaborative filtering algorithm,the user similarity calculation method did not consider the preference of users to the attributes of scheme and the different weights of attributes from different configuration perspectives.For this problem,an improved Person cosine similarity algorithm was proposed considering the objective weights of attributes for different schemes.BP neural network was used to calculate the objective weight of each attribute under different schemes.Users' preference for scheme attributes was considered to calculate the user similarity,which could accurately predict the target user's rating of scheme and recommend the best scheme under the condition of incomplete cold start.The scheme recommendation process of air purifier product service system was taken as an example to verify the proposed method could improve the accuracy of user similarity calculation and scheme recommendation.

Key words: product-service system, collaborative filtering, back propagation neural network, cosine similarity, air purifier

摘要： 产品服务系统是企业为用户提供的个性化需求解决方案,为了快速准确的设计方案,提出了改进的协同过滤算法根据用户的个性化需求推荐符合其需求特征的方案。针对传统协同过滤算法中,用户相似度计算方法没有考虑用户对方案属性的偏好问题,忽略了不同配置角度下各属性权重不同,提出了考虑方案客观属性的改进Person余弦相似度算法。首先采用BP神经网络计算获得不同方案下各属性的客观权重,其次考虑用户对方案属性的偏好计算用户相似度,可以在不完全冷启动的情况下准确预测目标用户对方案的评分并推荐最佳方案。最后,以空气净化器产品服务系统的方案推荐过程为例,验证了所提方法可提高用户相似度计算的精度与方案推荐的准确性。

关键词: 产品服务系统, 协同过滤, BP神经网络, 余弦相似度, 空气净化器