基于数据转移状态的性质判断模型研究

Research on Nature Judgment Model Based on Data Transfer Status

数据拥有的性质特点决定了其特殊的状态分布区间,研究数据的时空状态和性质特点对智能计算和数据挖掘等研究具有重要的意义。数据转移状态的概念描述了事物在特定的情况下从一种状态转化为另一个状态的概率,它能很好地反映数据不同状态之间协调比例情况。虽然目前的研究集中在使用肉眼直观数据对事物性质的判断上,但几乎没有关于数据转移状态确定事物性质的讨论。本工作考虑将数据转移状态应用于对事物性质分类的工作中,此外,本工作还引入了一个注意力时空模型(注意力时空特征网络,AttSTFN)。AttSTFN主要由三个模块组成,它们分别是空间特征提取模块、时序卷积层和空间注意力池化机制。该模型已在真实数据集上进行了评估,并与一些典型的模型进行了比较。实验结果表明,AttSTFN可以很好地反映时空信息并对数据的性质进行高精度判断分类。

Features of nature possessed by data determine its special state distribution interval, and the study of the spatio-temporal states and nature features of data is of great significance for research such as intelligent computing and data mining. The concept of data transfer state describes the probability of something transforming from one state to another in a specific situation, and it can well reflect the coordination ratio between different states of data. Although current research has focused on the determination of the nature of things using visual intuitive data, there has been little discussion on classifying the nature of things by data transfer states. In this work, we consider data transfer states for determining the nature of things in classification work, in addition, an attentional spatio-temporal model (attentional spatio-temporal feature network, AttSTFN) is introduced in this work. AttSTFN consists of three main modules, which are spatial feature extraction module, temporal convolutional layer and spatial attentional pooling mechanism. The model has been evaluated on a real dataset and compared with some typical models. The experimental results show that AttSTFN can well reflect the spatio-temporal information and classify the nature of the data with high accuracy judgment.