Abstract: Faced with the rapid growth of knowledge and the gradual formation of its dominant position in value creation, as a basic knowledge service method, knowledge push can effectively solve the problem of being lost in the direction of knowledge and accelerate the realization of knowledge value. A relevant knowledge push algorithm based on three-way decision was proposed based on a review of the existing recommendation techniques, and taking into account the importance of knowledge relevance, the dichotomy of traditional push decisions, and the cost of  knowledge push. Specifically, the association rule algorithm was used to mine the potential relationship between knowledge, and the knowledge relevance model which incorporates the knowledge attribute similarity and knowledge relevance was constructed, whose score could be predicted accordingly. In addition, in order to reduce the cost of misclassification in the traditional dichotomy, three-way decision ideas were introduced into the final recommendation decision to realize three-way push. Moreover, the superiority of the proposed knowledge push algorithm in terms of push effect and cost was verified by conducting a comparison experiment based on the Movielens datasets.

Key words: knowledge push, relevant knowledge, three-way decision, pushing cost

摘要： 面对知识体量的快速增长及其在价值创造中主体地位的逐渐成型，知识推送作为一项基础型知识服务方式，可以有效解决“知识迷向”问题，促进知识价值实现。在回顾现有推荐技术的基础上，兼顾知识关联度的重要性、传统推送决策的二分性质以及推送成本问题，提出了一种基于三支决策的关联知识推送算法。采用关联规则算法挖掘并利用知识之间的潜在联系，构建了融合知识属性相似度和知识关联度的知识相关度模型，并据此进行评分预测；同时，为了降低传统二分法中因错误分类而产生的成本，在最终的推荐决策中引入了三支决策思想，实现三支推送。最后，基于Movielens数据集，通过对比实验验证了本文知识推送算法在推送效果及成本方面的优越性。

关键词: 知识推送, 关联知识, 三支决策, 推送成本