基于群体协同智能聚类的期刊分类大数据自动存储方法

求解期刊分类大数据自动存储问题时,传统方法在分解的过程中无法保证准确性与合理性,对解的合并策略选择不合理,导致寻优过程中出现一定的偏差,造成期刊分类存储效率大大降低。为此,需要提出一种新的基于群体协同智能聚类的期刊分类大数据自动存储方法。确定径向基神经网络的初始结构,通过样本分布计算径基宽度获取隐节点群,将其当成初始集合。将分类存储精度最高、F-measure最大、期刊特征相似性最高作为目标函数,将其加权和作为适应函数。在求解过程中,各子群内部通过模拟退火法将分布估计算法和遗传算法结合在一起,产生新个体,利用群体协同合作的方式实现智能聚类。通过进化获取最优个体,得到最终精英集合,将其看作最后得到的径向基神经网络结构,通过得到的径向基神经网络实现期刊分类大数据自动存储。实验结果表明,所提方法期刊分类大数据存储性能强。

Automated Storage Method for Big Data in Periodical Classification Based on Group Collaboration and Intelligent Clustering

The traditional method of automating the storage of big data on periodicals cannot guarantee accuracy and rationality in the process of decomposition.A merging strategy is unreasonable, leads to deviations in the process of optimization, and greatly reduces the efficiency of the classification and storage of the periodicals.To solve this problem, we propose a new method of automating the storage based on swarm collaborative intelligence clustering.First, we determine the initial structure of a radial basis function neural network.Next, we obtaine the hidden node group by calculating the diameter of its base through a sample distribution and regarde it as the initial set.The highest classification storage accuracy, the largest F-measure, and the highest similarity of journal features are taken as objective functions, and weighted sums are used as fitness functions.In the process of solving, we combine a distribution estimation algorithm and a genetic algorithm to produce new individuals through simulated annealing and realized intelligent clustering by group cooperation.The optimal individual is obtained by evolution, and the final elite set is obtained.The periodical classification of big data is automatically stored by the obtained radial basis neural network.Experimental results show that the proposed method has strong storage performance.