基于深度学习的小目标检测区域数据推荐算法

针对电网数据推荐结果未考虑电网领域知识关联度,推荐结果精准度低,不能精准掌控电网工程建设管理工作问题,提出一种基于深度学习的小目标检测区域数据推荐算法。利用颜色直方图距离和边缘距离展开超像素合并,检测电网小目标区域,采用多尺度卷积神经网络特征提取方法获取小目标区域卷积神经网络特征。以获取的特征为知识项,通过知识树结构组织不同粒度知识项,构建电网领域知识体系关联和推理机制,采用协同过滤推荐算法AR-Item CF,挖掘用户行为日志,根据用户推荐列表,计算出不同知识项间的深层推荐。实验结果表明,该算法可有效解决推荐结果关联度低问题,且推荐效率高、质量好。

Data recommendation algorithm of small target detection area based on deep learning

Aiming at the problem that the results of power grid data recommendation do not consider the relevance of power grid domain knowledge,the accuracy of recommendation results is low,and the management of power grid engineering construction cannot be accurately controlled,a data recomm-endation algorithm based on deep learning for small target detection area is proposed.The super-pixel combination of color histogram distance and edge distance is used to detect the small target area of power grid,and the multi-scale convolution neural network feature extraction method is used to obtain the convolution neural network features of small target area.Based on the characteristics of knowledge items,different granularity knowledge items are organized through the knowledge tree structure,and the association and reasoning mechanism of knowledge system in power grid field is constructed.Ar-item-CF,a collaborative filtering recommendation algorithm,is used to mine the user behavior log,and the deep recommendation among different knowledge items is calculated according to the user recommendation list.The experimental results show that the algorithm can effectively solve the problem of low relevance of recommendation results,and the recommendation efficiency is higher and the quality is good.