基于改进后K-Means下网络自媒体空间数据挖掘

Network Self-Media Spatial Data Mining Based on Improved K-Means

摘要: 在手机、平板电脑等电子媒介的人均持有率大于一的今天，网络自媒体的传播达到了前所未有的巅峰。本文通过基于Hadoop平台的mahout数据挖掘框架，选用经过Canopy算法优化后的K-means聚类分析算法，对数据进行聚类分析，对内涵众多信息的网络自媒体推文进行数据挖掘，以发现微博数据中蕴含的与当下社会和生活相关的热点词，后通过ArcGIS，对文本类簇进行核密度分析，再做渔网栅格化分析，使离散的类簇样本具备邻接性，能在可视化中直观地看到类簇主题的主要分布情况，以研究人们日常生活中的习惯、了解单一个人的喜好，以及对某个社会事件的评价等日常生活中隐含着关于社会和生活相关的信息。

Abstract: Today, the percapita holding rate of electronic media such as mobile phones and tablet computers is greater than one. The spread of online media has reached an unprecedented peak. Based on the Mahout data mining framework of Hadoop platform, this paper selects the K-means clustering analysis algorithm optimized by Canopy algorithm, clusters the data, and mines the network self-media tweets with much information to discover the micro: the hot words related to the current society and life contained in the Weibo data, and then through the ArcGIS, the kernel density analysis of the text clusters, and then the fishing grid rasterization analysis, so that the discrete cluster samples have the adjacency, enabling visualization visually see the main distribution of cluster topics, to study the habits of people's daily lives, to understand the preferences of a single individual, and to evaluate the social events, such as social and life related information.