传感器网络环境下数据挖掘关键技术的研究

项目名称： 传感器网络环境下数据挖掘关键技术的研究

项目编号： No.60803015

项目类型： 青年科学基金项目

立项/批准年度： 2009

项目学科： 矿业工程

项目作者： 郭龙江

作者单位： 黑龙江大学

项目金额： 18万元

中文摘要： 传感网技术广泛应用于水资源监测等应用领域。网络中每个节点产生高速、实时的感知数据流。在这个分布式数据流中含有丰富的知识。获得这些知识，可以为用户做重大决策提供可靠的依据。但由于感知器件的质量及环境的影响，感知数据中存在大量噪声。另外，网络传输数据的不确定性以及感知器件的故障，容易造成感知数据缺失，这会使挖掘结果不精确。另外，节点硬件资源有限、能量有限、通信能力有限、网络拓扑经常动态变化等，这些问题是传感网中的数据挖掘面临的挑战。本项目提出一套有关传感器网络感知数据挖掘的理论、技术和方法包括：感知数据网内清洗、网内数据聚类、用于隐式防盗的网络内信号强度分类、网内近似频繁项集、用于移动传感网数据收集和移动目标跟踪中的预测技术、支持最小化聚集时间延迟的随机拓扑特性的新知识挖掘以及其他关键技术(网内区域数据查询、MAC协议，信道分配算法，簇头选择，睡眠调度机制)，在实验中验证了正确性，具有重要的学术意义和应用价值。

中文关键词： 传感器网络; 数据挖掘; 数据清洗

英文摘要： Wireless sensor networks have been broadly used to many application fields such as water resources monitoring. Each node in WSNs generates high speed and real time sensory data streams. There is rich knowledge in this distributed data streams. To achieve all these knowledge can provide reliable gist for users to making important decisions. Due to the effect of the quality of sensing components and environment changing, sensory data contains a lot of noise. In addition, some important part of data may be lost due to same reason. All above thoese reasons will result in imprecise mining results. Each node in WSNs has low computing ability, limited energy, limited communication ability. Network topology is volatile and frequent changing. All these issues become chanllenges in data mining in WSNs. This project proposes a set of theory, techniques and methods including sensory data cleaning in networks, data clustering in networks, classification of signal intensity for guarding against theft, approximate frequent item sets in networks, prediction technique for data collection in mobile sensor networks, new knowledge of mining random topology characteristics for supporting minimum aggregation latency and other key techniques (such as area query in networks, MAC protocols, channel assignment, cluster head selection, sleep scheduling). Correctness and validity are validated in simulations and real experiments. All those results have academical significance and application values.

英文关键词： sensor networks; data mining; data cleaning