基于聚合度热点适应机制的网络舆情 大数据收敛算法

为解决当前网络舆情大数据收敛算法普遍存在的收敛困难及热点聚类生成速度较低等难题,提出了一种基于聚合度热点适应机制的网络舆情大数据收敛算法。首先,通过增量用户节点与存量热点之间的信息交互关系,设计了一种基于聚合度初始化机制的数据收敛方案,采用匹配机制逐个对存量热点与增量用户节点间差异度及聚合度进行比对,能够将增量用户节点纳入性能最佳的存量热点所形成的种子聚类,提高聚类形成速度。随后,针对热点数量处于密集状态等极端情况,特别是用户特征匹配过程中难以实现快速匹配等不足,设计迭代方式,以逐步消除种子聚类差异度,提升大数据匹配性能,改善用户节点与热点之间信息交互质量。仿真实验表明:与当前常用的时间片累积挖掘收敛方案(Convergence Scheme for Time Slice Cumulative Mining,TSCM算法)及热点度显影收敛方案(Convergence Scheme of Hotspot Degree Development,HDD算法)相比,本文算法具有更高的收敛速度和聚类形成质量。

THE CONVERGENCE ALGORITHM FOR LARGE DATA OF NETWORK PUBLIC OPINION BASED ON HOTSPOT ADAPTATION MECHANISM OF CONVERGENCE DEGREE

In order to solve the common problems of convergence and low speed of hot spot clustering in current large data convergence algorithms of network public opinion, a new convergence algorithm of large data of network public opinion based on hotspot adaptation mechanism of convergence degree is proposed. Firstly, through the information interaction between incremental user nodes and stock hotspots, a data convergence scheme based on aggregation degree initialization mechanism is designed. The difference degree and aggregation degree between stock hotspots and incremental user nodes are compared one by one by using matching mechanism, which can bring incremental user nodes into seed clustering formed by stock hotspots with the best performance and greatly improve the aggregation class formation speed. Subsequently, aiming at the extreme situation that the number of hot spots is in a dense state, the difference degree of seed clustering is gradually eliminated by iteration, so as to improve the convergence performance of large data and the quality of information interaction between user nodes and hot spots. The simulation results show that, compared with the convergence scheme for Time Slice Cumulative Mining (TSCM) and the Hotspot Development convergence scheme (HDD), the proposed algorithm has the advantages of fast convergence speed, high clustering quality and strong practical deployment value.