基于机器学习的交叉覆盖信息捕获算法研究

Research on Cross Coverage Information Acquisition Algorithm Based on Machine Learning

随着系统复杂程度的增加,信息采集时很容易受到其它因素的影响而形成交叉覆盖,使信息具有不平衡、非线性等特点,难以实现信息的准确可靠捕获。为此提出一种基于机器学习的交叉覆盖信息捕获算法。为防止训练过程中单纯的距离计算引发邻近关系误判,采取欧式距离和局部均值相结合搜索特征属性的邻近元素,通过与特征属性的内外映射筛除原始信息中的非关联属性。为降低信息冗余度,利用信息熵描述未知信息量,并在信息熵基础上引入互信息来描述数据间的依赖关系,根据互信息矩阵完成主成分特征提取。最后利用机器学习的良好逼近性,构建SLFN学习网络,对网络模型进行正则化处理,并通过构建Lagrange函数求解网络输出加权,从而实现数据分类。仿真结果表明,所提算法对于不同复杂度的数据集具有更好的适应性,能够显著提高交叉覆盖信息捕获的准确率、鲁棒性,以及抗噪性。

With the increase of the complexity of the system, it is easy to form cross-coverage due to the influence of other factors, it makes information unbalanced and nonlinear, and it is difficult to capture information accurately and reliably.Therefore, a cross-coverage information acquisition algorithm based on machine learning is proposed.In order to prevent the simple distance calculation in the training process from causing the misjudgment of neighborhood relations, the Euclidean distance and local mean were combined to search the adjacent elements of feature attributes, and the nonassociated attributes in the original information were filtered out by the internal and external mapping with the feature attributes.In order to reduce the information redundancy, the information entropy was used to describe the unknown information.Based on the information entropy, mutual information was introduced to describe the dependence between data, and the principal component feature extraction was completed according to the mutual information matrix.Finally, the SLFN learning network was constructed by using a good approximation of machine learning, and the network model was regularized.And by constructing the Lagrange function to solve the network output weighting, the data classification was realized.Simulation results show that the proposed algorithm has better adaptability to data sets with different complexity, and can significantly improve the accuracy, robustness and noise resistance of cross-coverage information acquisition.