一种动态加权条件网络数据的特征标记算法

Feature Marking Algorithm Based on Dynamic Weighted Conditional Network Data

针对传统特征标记方法在面对海量的网络数据时出现的定位目标信息困难、时间和空间开销较大等问题,提出基于加权遗传算法的互信息特征反馈标记方法。首先优化数据处理流程,对目标数据特征进行加权处理,得到近似全局最优解;其次用户对文本特征或者图像实例完成标记,基于用户的标记与未标记情况构建双重监督图;最后建立实数值推测函数并计算,获取双重监督图中未标记的结点。通过仿真实验结果,验证了方法误差较小、检索精度较高,能够实现在大量的数据中快速找到目标内容。

Aiming at the problems that difficulties of locating target information,large time and space costs of traditional feature marking methods in the face of massive network data,a mutual information feature feedback marking method based on Weighted Genetic Algorithm is proposed.In this method,the data processing flow is optimized first,and the target data features are weighted to obtain an approximate global optimal solution.Then,the users mark text features or image instances,and a double supervision chart is constructed based on the user's marked and unmarked conditions.Finally,a real value speculation function is established and calculated to obtain unmarked nodes in the double supervision chart.The simulation results verify that the method has small error and high retrieval accuracy,and can quickly find the target content in a large amount of data.