多层感知学习下多维大数据智能分类检索仿真

Multi-Dimensional Big Data Intelligent Classification Retrieval Simulation under Multi-Layer Perception Learning

针对传统的多维大数据智能分类检索方法存在分类检索响应时间过长、误差较大、效率较低等问题,提出一种基于多层感知学习的多维大数据智能分类检索方法。将样本按照其特征进行分组,计算每组样本集的核矩阵,引入基于半正定规划的支持向量机(SVM)模型加入核矩阵线性组合,采用半正定规划SVM求解各个子特征空间的权重系数,利用权重系数计算特征贡献度以及支持度,用于特征分类。在分类结果中引入节点索引过滤多维大数据流程模型库,得到候选模型集,通过动态构建的边索引对候选模型集进行二次过滤,将该算法与并行算法相结合,通过计算结果完成多维大数据智能分类检索。根据实验结果看出,文中所述方法可以有效有效减少了多维大数据分类检索时间,降低分类检索误差,提高了分类检索效率。

Traditional method leads to long response time, large error and low efficiency of classification retrieval. Therefore, an intelligent method of classification retrieval for multi-dimensional big data based on multi-layer perception learning was put forward. Firstly, the samples were grouped according to the characteristics, and the kernel matrix of sample set in each group was calculated. Secondly, the support vector machine model based on semi-definite programming was introduced into the kernel matrix linear combination, and then the semi-definite programming SVM was used to solve the weight coefficient of each sub-feature space. Moreover, the weight coefficient was used to calculate the feature contribution degree and the support degree, which could be used for the feature classification. In the classification result, the node index was introduced to filter process model base of multidimensional big data, so that the candidate model set was obtained. Then, the candidate model set was filtered by the edge index that was dynamically constructed. Finally, this algorithm was combined with the parallel algorithm. Thus, the multidimensional data intelligent classification retrieval was completed. Simulation results show that the proposed method effectively reduces the multi-dimensional big data classification retrieval time. Meanwhile, this method reduces the classification retrieval error and improves the classification retrieval efficiency.