一种交互网络特征反馈标记方法研究

Research on a feature feedback labeling method for interactive network

面对海量的网络数据,传统方法在检索信息时需要庞大的精力和时间,因此,提出基于加权遗传算法的交互网络特征反馈标记方法。分析交互网络数据处理流程,根据分析结果,利用加权遗传算法对特征加权计算,找出近似全局最优解;使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,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。仿真实验验证了方法误差较小、检索精度较高,可在大量的数据中快速找到目标内容。

When facing with massive network data,traditional methods need to pay a lot of energy and time to retrieve information.Therefore,this paper proposes an interactive network feature feedback labeling method based on weighted genetic algorithm.The whole process of data processing of interactive network is analyzed.According to the results,the weighted genetic algorithm is used to calculate the weight of features to find the approx-imate global optimal solution.After that,users can mark the text features or image instances,and the algorithm could build a double supervision graph based on the marked and unlabeled situation of users,and then a real value inference function is built to calculate the unlabeled nodes in the double-supervised graph.The simulation results show that the method has small error and high retrieval accuracy,and can quickly find the target content in a large amount of data.