大数据驱动下基于深度学习的药品信息分布式检索
Distributed Retrieval of Drug Information Based on Deep Learning Driven by Big Data
单增朗杰1
作者信息
- 1. 西藏自治区药品监督管理局,西藏拉萨 850000
- 折叠
摘要
由于药品信息自身构成较为复杂,导致检索结果的相关性难以得到保障,为此,提出大数据驱动下基于深度学习的药品信息分布式检索研究.通过设计一个针对药品信息特征匹配的神经网络判别器,开展药品信息匹配分析,并引入了交叉熵损失函数,对二分类问题中(判断两个药品是否匹配)的真假进行计算;在分布式遍历检索阶段,为节点分配一个唯一的标识符,并设置相应的访问状态为未访问,通过将"最大遍历节点数"作为变量,对检索的遍历过程中进行动态更新,在先进先出原则下,当所有节点都被访问过,并且没有新的药品信息被检索到时,遍历过程结束,输出与目标检索信息拟合度最高的TOP-N药品信息队列.在测试结果中,检索结果的相关性始终在0.90以上,最大值为0.947.
Abstract
As the composition of drug information itself is relatively complex,it is difficult to guarantee the relevance of search results.Therefore,a research on distributed retrieval of drug information driven by big data based on deep learning was proposed.A neural network discrim-inator for matching drug information features was designed to carry out drug information matc-hing analysis,and cross entropy loss function was introduced to calculate the true and false of the binary classification problem(judging whether two drugs match).In the distributed traversal retrieval stage,a unique identifier is assigned to the node,and the corresponding access status is set as unvisited."Maximum number of traversal nodes"is taken as a variable to dynamically update the traversal process of retrieval.Under the principle of FIFO,when all nodes have been accessed and no new drug information is retrieved,the traversal process ends.Output TOP-N drug information queue with the highest fit to the target retrieval information.In the test re-sults,the correlation of the search results is always above 0.90,and the maximum value is 0.947.
关键词
深度学习/药品信息/分布式检索/特征匹配/神经网络判别器/交叉熵损失函数/先进先出原则Key words
Deep learning/Drug information/Distributed search/Feature matching/Neural net-work discriminator/Cross entropy loss function/First in first out principle引用本文复制引用
出版年
2024
NETL
NSTL
万方数据