Research On Data Dimension Reduction And Engineering Application Based On Manifold Learning

With the rapid development of artificial intelligence and large data technology,machine learning and data mining of these artificial intelligence core research areas tend to be generally showing high dimensional,non-linear characteristics of the data.Taking analog circuit fault diagnosis as an example,especially in a highly integrated circuit board for board-level fault location,due to the large number of components and the existence of tolerance,resulting in the data collected more inclined to high-dimensional,non-linear Structure distribution.For these large-scale,high-dimensional non-linear data,people want to intuitively perceive useful knowledge hidden in high-dimensional data,the difficulty can be imagined.Data dimension reduction is one of the most effective ways to reduce the dimensionality of high dimensional data.Data dimension reduction technology can be divided into linear dimensionality reduction and nonlinear dimensionality reduction.The linear dimensionality reduction technique is widely used,but the effect of reducing the dimensionality of the actual engineering application with large amount of non-linear data is also becoming the hotspot of the current research.In order to obtain the low-dimensional representation of high-dimensional and difficult to understand data,the local linear embedding dimension reduction method based on manifold learning uses the assumption of local linear and global nonlinearity to reduce the dimension of high dimensional data,Can still maintain the original structure of high-dimensional data.This feature makes it one of the hotspots in the field of machine learning.In this paper,we study the problem of feature reduction and feature extraction using Local Linear Embedding(LLE)technique based on manifold learning method.Aiming at the problem of high feature dimension in analog circuit fault diagnosis project,a feature dimension reduction scheme based on Wavelet Packet Decomposition(WPD)and LLE algorithm is proposed.The two-stage four-op amp low-pass filter circuit is studied.Research on fault feature dimension reduction based on Clone Selection Algorithm(CSA).The experimental results verify the applicability of the algorithm proposed in this paper to reduce the problem of feature reduction in analog circuit fault diagnosis,which provides a useful reference for LLE algorithm to reduce the engineering of complex data.