基于商空间的不完备形式背景填补方法研究  

RESEARCH ON IMPUTATION APPROACH IN INCOMPLETE FORMAL CONTEXT BASED ON QUOTIENT SPACE

作　　者：张其文[1] 王培瑾 Zhang Qiwen,Wang Peijin(School of Computer and Communication, Lanzhou University of Technology, Lanzhou 730050, Gansu, Chin)

机构地区：[1]兰州理工大学计算机与通信学院,甘肃兰州730050

出　　处：《计算机应用与软件》2018年第8期37-42,49,共7页Computer Applications and Software

基　　金：国家自然科学基金项目(61363078)

摘　　要：目前,在常用缺失数据填补算法中,需要逐个计算所有对象之间的相似度。然而,在数据缺失率较高的情况下,相似度的计算难度也随之加大,导致填补效率低下、填补准确度不高。为提高算法效率和准确度,结合粒计算,提出基于商空间的数据填补算法。引入对象间的近似度上界、下界,计算模糊近似评估矩阵;通过分层递阶结构的推理模型,对对象集进行粒化,压缩了对象集规模,提高了算法效率;根据实际应用需求给出填补结果。通过在真实数据集上的实验表明,该算法在数据缺失率较高时能够显著提高填补效率,并且保持高准确性。At present, in the commonly used missing data imputation algorithm, the similarity between all objects needs to be calculated one by one. However, in the case of data missing rate, the computation difficulty of similarity is also increasing, resulting in low filling efficiency and low accuracy. In order to improve the efficiency and accuracy of the algorithm, combined granular computing, a data filling algorithm based on quotient space was proposed in this paper. The upper and lower bounds of approximation degree between objects were introduced, and the fuzzy approximate evaluation matrix was calculated. Through the hierarchical structure reasoning model, the object set was granulated, the object set size was compressed, and the algorithm efficiency was improved. The results were given according to actual application requirements. Experimental on real datasets show that the proposed algorithm can significantly improve the efficiency of filling and maintain high accuracy when the data loss rate is high.

关 键 词：缺失数据填补 商空间 模糊近似评估矩阵 分层递阶结构 粒 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]