物联网用户查询目标自动推荐算法仿真研究

Simulation of Automatic Recommend Algorithm for Users Query Target in Internet of Things

查询推荐作为一种改善物联网用户查询体验和效率的有效方式,可以帮助用户筛选并提供更加准确的查询目标描述。目前多数推荐算法随着用户数量和项目数量的增加,存在着推荐精度低、推荐时间长等问题,针对以上问题,提出基于改进近似性度量的物联网用户查询目标自动推荐算法。结合TF-IDF算法和信息熵算法构建物联网用户对项目属性的兴趣度模型,运用自组织映射神经网络聚类算法对兴趣度模型中的用户进行初步聚类。并以此为基础数据在目标用户所属簇内计算用户近似度并寻找最近邻。针对K-Means算法因缺少共同评分项目而导致近似度较低的弊端,采用改进近似性度量方法在簇内寻找目标用户的最近邻和项目推荐候选集,根据项目推荐候选集上的预测评分,完成用户查询目标自动推荐。实验结果表明,所提算法与其它算法相比,有更好的推荐精度和效率。

An automatic recommendation algorithm for query targets of users in Internet of Things based on improved similarity measures was presented. Combined TF-IDF algorithm with the information entropy algorithm, the interest model of users in Internet of Things on project attributes was constructed. Then, self-organizing mapping neural network clustering algorithm was used to conduct preliminary clustering for users in the interest model. Based on these data, the user approximation was calculated in the cluster of target users and the nearest neighbor was found. For the shortcomings that K-Means algorithm caused the low approximation degree due to lack of common scoring item, the improved similarity measure method was used to find the nearest neighbor of target user within cluster and the project recommendation of the target user in the cluster. Finally, the automatic recommendation of user query target was completed based on the prediction score of recommendation candidate set. Simulation results prove that the proposed algorithm has better recommendation accuracy and efficiency than other algorithms.