基于用户情景推断的军事信息Hybrid-CF推荐算法

Military Information Hybrid-CF Recommendation Method Based on User Scenario Inference

摘要:

针对当前军事信息推荐方法未对用户所处时间、地理位置和用户场景做区分,以及未考虑军事用户的信息需求与所处场景的关联性,导致推荐结果固定、单一化问题,设计了基于用户情景推断的军事信息Hybrid-CF(混合协同过滤)推荐方法；融入军事用户情景要素对传统协同过滤算法进行了改进,通过计算当前情景信息与历史信息的相似度,更加准确地推断出当前军事用户的所处情景,继而给军事用户推荐符合其需求的特定情景下的军事信息；为了解决推荐算法矩阵稀疏、效果单一等问题,引入了加权平衡因子将不同的推荐算法进行动态加权得到融合情景信息的Hybrid-CF推荐算法,并通过控制因子λ对加权平衡因子进行动态调整；实验结果表明,所提出的Hybrid-CF推荐算法在准确率和召回率上均体现了良好的提升效果。

Abstract:

Aiming at the problem that the current military recommendation algorithm does not distinguish between the time, geographical location and user scenario, and the disregard of the relevance of military users’ information needs and scenario, resulting in fixed and single recommendation results, the authors designs Military Information Hybrid Collaboration Filter Recommendation Method Based on User Scenario Inference. Through integrating the scenario information, the collaborative filtering algorithm has been improved and the similarity between the current scenario information and the historical information has been calculated, so as to accurately infer the scenario of the military user and provide the recommendation service for the military user in a specific scenario. In order to solve the problem of sparse matrix and single effect of recommendation algorithm, the authors introduces a weighted balance factor to add different recommendation algorithms to a hybrid recommendation algorithm that dynamically weights the fusion scenario information and the controlling factor λ is introduced for dynamic adjustment. Experimental results show that the accuracy and recall of hybrid recommendation algorithm recommendations are improved.