基于数据流的时间条件占优查询

Query time⁃conditional preference query based on data flow

传统的偏好推理使用权衡增强的条件偏好网络(Tradeoff-Enhanced Conditional Preference Networks,TCP-nets)进行用户的偏好推理,不仅能高效地表示对元组的定性偏好关系并优化用户偏好结果,还能描述每个属性之间的偏好关系,其主要聚焦于关系元组中的单个属性的偏好.但把对条件偏好查询的技术推广到数据流的条件提取却是一个挑战,面临的技术困难主要是对数据流中序列的提取,对提取的序列进行占优查找等.首先,针对偏好数据流,提出一种时间条件查询语言Stream Pref来处理数据流;其次,在Stream Pref中加入时间索引来推理和规范数据流提取序列的时间条件偏好,提出提取对象序列算法、占优对象及占优序列查找算法和数据流序列间占优对比的算法;最后,在数据集上分析验证提出的算法的有效性.实验结果证明,提出的算法与min Top-k,Partition和Incpartition算法相比,得到的结果更准确.

Traditional preference inference uses tradeoff⁃enhanced conditional preference networks for user preference inference,which not only efficiently represent qualitative preference relations over tuples and optimize user preference results,but also describe preference relations between each attribute.The main focus is on the preference of individual attributes in relational tuples,but it is a challenge to extend the technique of conditional preference query to the conditional extraction of data streams,and the technical difficulties are mainly the extraction of sequences in the data streams and the preference finding of the extracted sequences.Firstly,a temporal conditional query language Stream Pref is proposed to process the data streams for preference data streams.Secondly,Stream Pref incorporates a temporal index to reason and standardize the temporal conditional preferences of the extracted sequences of data streams.An algorithm for extracting object sequences,an algorithm for finding preference objects and preference sequences and an algorithm for preference comparison among data stream sequences are proposed.Finally,the effectiveness of the algorithm proposed in this paper is analyzed and verified on the data set.Experimental results show that the proposed algorithm gets more accurate results compared with min top⁃k algorithm,partition algorithm and incpartition algorithm.