Dimension Reduction Of Industrial Monitioring Data And Its Application

Dimension reduction is the most critical parts in the high-dimensional data analysis.With the coming of the era of artificial intelligence,the data which containing valuable resources is growing in a blowout.How to extract effective information from mass industrial monitoring data becomes a core problem,the problem is ubiquitous because the higher spatial dimension leads to the strong sparsity of data and the difficulty of matrix decomposition,which makes it more complex t.o analyze the data.To solve the above problems,this paper proposes a step by step framework of dimension reduction based on the study of the traditional dimension reduction algorithm.At the same time,the incremental orthogonal component analysis algorithm was improved,and the simulation proved that the framework of dimension reduction and the improved algorithm has a higher performance.The research in this paper includes the following:1 Learning several basic dimension reduction algorithm and its improved algorithm,and realizing the algorithms.The last it through the visual space to observe the efifect of algorithm of different dimension reduction;2?In this paper,a step by step framework of dimension reduction based on clustering was proposed,and the relevant model was established by efficiently with the combination common clustering algorithm and dimension reduction algorithm.The effect of framework was analyzed.It is used for linear and non-linear high-dimensional data,and the simulation results indicate that framework improved the accuracy of classification.The classification accuracy can be improved by 16.9%under certain conditions.3?The proposed step by step dimension reduction framework was improved,and the dimension of data was presorted in the early stage of the model processing,and instead of selecting the parameters of the clustering algorithm through prior knowledge.The improved model was used to recover the missing data of train axle temperature,The effectiveness of the proposed framework and the situation of data recovery was analyzed from the time and recovery precision,and the validity of the dimension reduction framework was verified.The simulation result shows that the time eff-iciency was improved by 14.25%when the error accuracy is guaranteed,and the average accuracy of multi-dimensional missing data can be as high as 99.75%.4?In this paper,the incremental orthogonal component analysis algorithm was improved,and from the perspective of probability theory,the cumulative distribution curve of beta was selected as the adaptive threshold function of the algorithm,and the online model of dimension reduction was established.Schmitt which the feature space to meet the nature of standard and orthogonal was introduced by the model.Finally the validity of algorithm by simulation experiment,it can adjust the parameters of the adaptive thresholding function to update the target dimension,and the same time it find a balance between the two.It is important that it provides an effective method and means to deal with high-dimensional dynamic data.