面向推荐系统的多目标决策优化算法

A Multi-Objective Decision Optimization Algorithm for Recommendation System

针对推荐系统利用多目标决策技术进行位置信息的查询与推荐时,由于查询者位置的移动和空间障碍物的位置变化导致传统多目标决策技术的查询效率较低的问题,提出了一种基于范围的障碍空间连续Skyline查询算法。首先,根据静态Skyline点的特征对由空间数据对象信息组成的初始数据集进行约减;然后,根据障碍空间中查询者的位置移动的特点构建距离相交模型,利用距离相交模型和数据对象的属性提出了剪枝策略,再根据剪枝策略过滤掉当查询者的位置移动时对查询结果无影响的数据对象,从而精减了冗余数据,得到过滤后的候选数据集;最后,根据数据对象的非空间属性和相互间的支配关系特征得出影响候选数据集的事件,利用影响候选数据集的事件再对候选数据集进行精炼计算,从而减少了冗余计算,查询出当前时刻的结果集。理论研究与实验结果表明:所提算法在查询者位置移动和空间障碍物位置变化时,能提升多目标决策技术的查询效率;相对传统对比算法,在数据集规模、障碍物数量、查询范围增大时,所提查询算法的平均效率提升约13%;针对多维度数据信息的查询,所提查询算法的平均效率提高了约11%。

This paper proposes a continuous range Skyline query algorithm in obstacle space to solve the problem of low query efficiency with traditional multi-objective decision technology caused by the movement of searcher’s position and the change of spatial obstacles in the recommendation system.Firstly,the initial data set composed of object information in the spatial data is reduced according to the characteristics of the static skyline points;then,the distance intersection model is constructed according to the characteristics of the position movement of the querier in the obstacle space,and the pruning strategy is proposed by using the distance intersection model and the attributes of the data objects.According to the pruning strategy,the data objects that have no impact on the query results when the position of the querier moves are filtered out,so as to reduce the redundant data and obtain the filtered candidate data set;finally,according to the non-spatial attributes of the data object and the characteristics of the dominant relationship between them,the events affecting the candidate data set are obtained and used to refine the candidate data set,so as to reduce redundant calculation and query the result set at the current time.Theoretical research and experiments show that the proposed algorithm can improve the query efficiency of multi-objective decision technology when the searcher moves and the position of spatial obstacles changes.Compared with the traditional comparison algorithm,the average efficiency of the query algorithm is improved by about 13%when the size of the data set,the number of obstacles and the query range are increased;for the query of multi-dimensional data information,the average efficiency of the proposed query algorithm is improved by about 11%.