海量稀疏大数据混合特征个性化推荐算法仿真

Simulation of Massive Sparse Big Data Mixed Feature Personalized Recommendation Algorithm

在大规模集群稀疏数据中使用混合特征技术处理海量大数据,可以显著优化推荐算法的可扩展性。于是设计了量稀疏大数据混合特征个性化推荐算法。利用逻辑运算方法处理稀疏数据,获取数据之间的关联性系,并及时填补缺失数据。针对大数据具有的数值和属性两种混合特征,通过计算相异性测度,建立混合特征聚类目标函数,实现数据聚类。基于聚类结果,从登陆、注册、检索浏览习惯等方面采集用户行为特征信息。采用显著数据分区检测方法融合用户信息,建立用户偏好挖掘模型,以行为偏好为基础,计算用户对内容的评分情况,将所有项目按照评分值排序,生成推荐列表。仿真结果表明,研究方法的同类大数据聚类准确度更高,平均绝对误差低于0.04,验证了上述方法的推荐结果可满足用户需求。

Using mixed feature technology to process massive amounts of big data in large-scale cluster sparse da⁃ta can significantly optimize the scalability of recommendation algorithms.So a personalized recommendation algorithm for sparse big data mixed features was designed.Firstly,the logical operation method was used to process sparse data and thus to obtain the correlation between data.Meanwhile,the missing data were filled in time.Aiming at the mixed characteristics of big data,namely the value and attribute,the clustering objective function of mixed feature was es⁃tablished by calculating the dissimilarity measure,so that data clustering can be completed.Based on the clustering results,the user behavior characteristics were collected from the aspects of login,registration,retrieval and browsing habits.Moreover,user information was fused by the method of significant data partition detection.Then,a user pref⁃erence mining model was built.Based on users'preference,user ratings for content were calculated.All items were sorted according to the scoring values.Finally,a preference list was formed,and the user preference was calculated.Simulation results show that the proposed method has higher clustering accuracy for similar big data;the mean abso⁃lute error is less than 0.04,so the recommendation result can meet the users'demands.