因素空间理论的因素约简算法

Factor reduction and data mining based on factor space

为讨论差转计算在多因素决策问题中对因素系统的约简性能,通过理论分析与实证检验,在算法原理与机制的基础上深入讨论了算法的因素约简能力,并与基于差别矩阵的粗糙集因素约简算法进行比较.研究结果表明:差转计算在进行决策的过程中,确保了对结果有重要影响的因素进入经验推理系统,影响不大的因素在决策的过程中自动地被舍弃,从而实现了因素的约简.在6个UCI数据集上,通过对因素约简结果与时间复杂度两个方面的讨论,得出差转计算的因素约简能力同差别矩阵算法相当,在时间复杂度方面远优于差别矩阵算法.

In order to discuss the effectiveness of factor system reduction of Subtraction Rotation in multi-factor decision problems, by theory analysis and demonstration test, this paper discussed the ability of factor reduction in depth on basis of algorithm principle and mechanism, and then compared it with rough set algorithm based on different matrix. The results show that in the decision process, Subtraction Rotation is sure of putting factors that have important effect on experience reasoning system, while automatically dropping factors that have little effect, thus achieves the reduction goal. In 6 UCI data sets, this study discussed the results and time complexity, and obtained the conclusion that the ability SR for factor reduction is equal to that of different matrix algorithm, but Subtraction Rotation is better in time complexity.