跳跃跟踪SSA交叉迭代AP聚类算法

Jump Tracking SSA Hybrid Iterative AP Clustering Algorithm

针对传统近邻传播聚类算法以数据点对之间的相似度作为输入度量,由于需要预设偏向参数p和阻尼系数λ,算法精度无法精确控制的问题,提出了一种跳跃跟踪麻雀搜索算法优化的交叉迭代近邻传播聚类方法.首先,针对麻雀搜索算法中发现者和加入者位置更新不足的问题,设计了一种跳跃跟踪优化策略,通过考虑偏好阻尼因子的跳跃策略设计大步长更新发现者,增加麻雀搜索算法的全局勘探能力和寻优速度,加入者设计动态小步长跟踪领头雀更新位置,同时,利用自适应种群划分机制更新发现者和加入者的比重,增加算法的后期局部开发能力和寻优速度;其次,设计基于扰动因子的Tent映射,在此基础上增加3个参数,使映射分布范围增大,并避免了陷入小周期点和不稳周期点;最后,引入轮廓系数作为评价函数,跳跃跟踪麻雀搜索算法自动寻找较优的p和λ,代替手动输入参数,并融合基于扰动因子的Tent映射优化近邻传播算法,交叉迭代确定最优簇数.使用多种算法聚类University of California Irvine数据集的10种公共数据集,仿真结果表明,本文提出的聚类算法与经典近邻传播算法、基于差分改进的仿射传播聚类算法、基于麻雀搜索算法优化的近邻传播聚类算法和进化近邻传播算法相比具有更优的搜索效率以及聚类精度.对国家信息数据进行了聚类分析,提出的方法更加准确有效合理,具有较好的应用价值.

Aiming at the problem that the traditional affinity propagation(AP)clustering algorithm takes the similari⁃ty between data points as the input measure,and the accuracy of the algorithm cannot be accurately controlled due to the need to preset the preference(p)and the damping coefficient(λ),a jump tracking sparrow search algorithm(JTSSA)opti⁃mized hybrid iterative AP clustering method(JTSSA-AP)is proposed.Firstly,in order to solve the problem of insufficient update of the position of the producers and the scroungers in sparrow search algorithm(SSA),a jump tracking optimization strategy is designed.By considering the preference factor,the jump strategy updates the producers in a large step,which in⁃creases the global exploration ability and optimization speed of SSA algorithm.The scroungers dynamically track the up⁃date position of the leading sparrow in a small step,and uses the adaptive population division mechanism to update the pro⁃portion of the producers and the scroungers,which increases the late local development ability and optimization speed of the algorithm.Secondly,on the basis of the original Tent mapping,the disturbance factor is added and three parameters are added,so that the mapping distribution range is increased and the small periodic point and unstable periodic point are avoid⁃ed.Finally,the silhouette index is introduced as the evaluation function,JTSSA is designed to automatically find better p andλinstead of manual input parameters,the tent map based on disturbance factor optimize the AP clustering,and the opti⁃mal number of clusters is determined by hybrid iteration.Multiple algorithms are utilized to cluster the 10 public datasets of the university of California Irvine dataset.Simulation results indicate that the proposed clustering algorithm in this paper exhibits superior search efficiency and clustering accuracy compared to the AP algorithm,the AP clustering algorithm based on differential evolution,the AP clustering algorithm optimized by SSA,and the evolutionary affinity propagation.Cluster analysis is conducted on country data,and the proposed method demonstrates greater accuracy,effectiveness,and rationali⁃ty,showcasing considerable practical value.