摘要
根据粒子群优化算法的思想,给出了求解高维邻域决策表的一个约简算法SPRA。通过采用固有维数的分析方法MLE等,将其估算的维数值作为SPRA算法的初始化参数,提出了高维数据集快速约简算法QSPRA。利用5个UCI标准数据集对该算法进行了验证,结果表明,该算法是有效的、可行的。详细分析了种群规模和迭代次数对结果产生的影响。实验表明,基于核的启发式添加算法思想已经不适合求解高维数据集。
According to the particle swarm optimization algorithm＇s idea,a new algorithm（SPRA） to get a optimal attribute reduction on the high-dimensional neighborhood decision table was proposed.Through the use of intrinsic dimension analysis method,taking the intrinsic dimensionality estimated as the SPRA algorithm＇s initialization parameter,a quick reduction algorithm（QSPRA） was proposed to deal with the high-dimensional data sets.The algorithm＇s validity was verified by five high-dimensional data sets from UCI.In the experimental analysis section,the population size and the number of iteration to the influence of the reduction result were also discussed.Moreover,the experiments also show that it is impossible to solve high-dimensional data sets based on kernel-based heuristic algorithm ideas.
出处
《计算机科学》
CSCD
北大核心
2012年第10期268-271,317,共5页
Computer Science
基金
国家自然科学基金(61170106)资助
关键词
邻域粗糙模型
决策依赖度
固有维数估算
极大似然估计法
粒子群优化算法
粒子群快速约简算法
Neighborhood rough set model
Decision-making dependency
Intrinsic dimension estimation
MLE
Particle swarm optimization algorithm
Quick particle swarm reduction algorithm
