基于ETL和SVM的融媒体平台数据采集与分析技术研究

Design of data collection and analysis technology for converged media platform based on ETL and SVM

针对融媒体平台中不同终端存在数据独立、技术不统一、难以实现数据共享的问题,文中利用ELT以及支持向量机技术开展了融媒体平台数据采集与分析方法的研究。通过构建数据中心,将不同终端、平台的数据接口统一定义,并采用“星型”结构的数据中心模式有效降低接口程序的数量,避免因子系统改变引发的数据结构变化。对于不同的业务需求,ELT技术可根据相关模型的需求筛选得到所需的数据子集,完成相应的特征提取和模型计算,从而降低对源端的依赖与访问频次,同时对于机器学习算法具有良好的兼容性。通过将聚类代理评价与支持向量机结合,对所有样本实例数据进行聚类初始化,在进化过程中对种群中的所有个体进行编码聚类。根据代理评价结果选择出进行SVM评价的个体,提高SVM模型的运行时间及精准度。测试结果表明,所提算法对数据采集、高用户访问具有较高的稳定性,平均分类精度为78.5%。与其他算法相比,其分类精度较高。

Aiming at the problems of data independence,technology inconsistency,and difficulty in data sharing between different terminals in the converged media platform,ELT and support vector machine technology are used to carry out the research on the data collection and analysis methods of the converged media platform.By building a data center,uniformly defining the data interfaces of different terminals and platforms,and adopting a"star"structure data center model can effectively reduce the number of interface programs and avoid data structure changes caused by changes in the factor system.For different business requirements,ELT technology can filter the required data subsets according to the requirements of related models,complete the corresponding feature extraction and model calculations,thereby reducing the dependence on the source and the frequency of access,and at the same time,it has good machine learning algorithms.compatibility.Through the combination of clustering agent evaluation and support vector machine,cluster initialization is performed on all sample instance data,and all individuals in the population are encoded and clustered in the evolution process.According to the results of the agent evaluation,the individuals for SVM evaluation are selected,which improves the running time and accuracy of the SVM model.The test results show that the proposed algorithm has high stability for data collection and high user access data.The average classification accuracy is 78.5%.Compared with other algorithms,its classification accuracy is higher.