基于机器学习的矿井三维空间无线传感器网络节点感知算法研究

Research on Mine 3d Space Wireless Sensor Network Node Perception Algorithm Based on Machine Learning

为了解决矿井无线传感器网络感知节点在保证网络系统能耗最小条件下被尽可能感知的问题,提出了一种基于机器学习的矿井三维空间无线传感器网络节点感知算法。首先对节点的历史监测数据信息进行挖掘;进行二维平面的映射和标准化处理,计算其属性个数。其次计算每个属性维度的熵信息权重系数,并利用节点密度分布函数计算各属性的影响因子。然后利用修正的特征节点距离计算公式和空间数据集的空间关系计算其节点的领域,依据节点的领域和密度阈值迭代的进行类的聚类分析。最后利用SVM算法对缩减后的样本集进行二次训练。实验仿真证明,该算法节点感知精度高、节点聚类训练速度快、计算量小、网络能耗最优。

In order to solve the mine energy consumption of wireless sensor network （WSN） nodes in the network system perception is perceived as possible under the condition of minimum problem. A coal mine based on machine learning perception algorithm for wireless sensor network nodes in three dimensional space is put forward. In the algorithm, first of all, the history of the node monitoring data information mining, for 2 d planar mapping and standardized processing, calculating the number of attributes. Secondly each attribute dimension information entro- py weight coefficient is calculated, and the node density distribution function is used to calculate the impact factor of each attribute. Then modified feature node distance calculation formula and spatial data set of spatial relations through the calculation of node areas, based on the node field and density threshold iteration of clustering analysis are used. Finally the SVM algorithm is used to shrink after secondary training sample set. Experiment results show that the algorithm nodes perception of high precision, node clustering the training speed, small amount of calcula- tion and energy consumption of network optimal.