基于多场景融合的分布式推荐模型

中文摘要: 个性化推荐系统中普遍存在着信息共享程度低、资源复用不足等问题。针对这些问题，提出基于多场景融合的分布式推荐模型，给出了该模型的组成单元和运行流程，以及对应的场景数据结构。该模型采用分布式的双向刻画的方法，通过多场景融合算法，进行客户特征（需求）与服务场景的互生成，并最终生成推荐列表。仿真实验证明，该模型较之独立节点的推荐模型，在消费娱乐领域，具有较高的客户覆盖度、推荐精度，且占用系统资源较少，具有较高的性价比。
Abstract:In order to deal with the lack of information sharing and resource re-utilizations,a novel personalized recommendation model was presented based on multi-scene fusion,and its function units,working flows and the scene data structure were given.The model utilized a distributed bidirectional description method to process scene data,and multi-scene fusion algorithms were used to exchange recommendation information between client characters and service scenes.Simulation results showed that the model has better coverage rates,recommendation precisions and resource consumption rate than the absolute node algorithms.