基于鱼群算法的脑功能连接邻域粗糙集特征归约方法

为了有效应对脑功能连接高维小样本性给分类模型构建带来的挑战，得到与脑疾病诊断相关的重要特征，提出基于鱼群算法的脑功能连接邻域粗糙集特征归约方法.该方法建立脑功能连接数据的邻域决策表；依据特征的依赖度将鱼个体初始化为候选的脑功能连接特征子集，并采用综合特征子集依赖度和特征子集长度的适应度函数对鱼个体进行评价；在种群优化过程中，执行觅食、聚集、追尾机制，以及交叉和迁徙2个新机制来不断搜索最优的特征子集.在3种脑疾病功能磁共振脑成像（fMRI）数据集上，将所提方法与多种已有的特征归约方法进行对比实验.结果表明，该方法是有效的脑功能连接特征归约方法，可以有效降低脑功能连接数据的维度，获得分类判别能力较强的脑功能连接特征.

Feature reduction of neighborhood rough set based on fish swarm algorithm in brain functional connectivity

Feature reduction of neighborhood rough set based on fish swarm algorithm in brain functional connectivity was proposed, in order to effectively deal with the challenge brought by the high-dimensional and small sample size of brain functional connectivity to the construction of classification model, and to obtain important features related to brain disease diagnosis. The neighborhood decision table of the brain functional connectivity is established in the algorithm. Each artificial fish is initialized as a candidate feature subset of the brain functional connectivity according to the feature dependence information, and a fitness function is constructed based on the information of feature subset dependence and feature subset length to evaluate each individual. The preying, swarming, following mechanisms of the fish swarm algorithm, as well as two new simulation mechanisms of crossover and migration are performed to iteratively search for the optimal feature subset in the process of population optimization. The proposed method was compared with a variety of existing feature reduction methods in the functional magnetic resonance imaging (fMRI) data sets of three brain diseases. Results show that the new method is an effective feature reduction method for the brain functional connectivity, which can effectively reduce the dimension of the brain functional connectivity data and obtain the brain functional connectivity features with high classification discrimination ability.