The local linear embedding (LLE) algorithm needs to manually specify the number of nearest neighbors when mining the local structure of high-dimensional space

which cannot guarantee the algorithm′s feature extraction ability. To address this issue

a local enhanced linear embedding algorithm with sample density adaptation (SDA-LELE) is proposed. Firstly

the sum of distances between the sample points and their nearest neighbors is used to measure the sparsity and density of the sample distribution

thereby adaptively selecting the number of nearest neighbors. Secondly

a local enhancement algorithm is adopted to increase the weight between the adjacent samples

so that the samples maintain both local linear structure and local nearest neighbor structure

enhancing the algorithm′s feature extraction ability. Finally

the algorithm is applied to the bearing data sets of Case Western Reserve University and Northeast Petroleum University

and visualization

Fisher information and other experiments are carried out. The experimental results show that the SDA-LELE algorithm can extract more significant features and achieve better dimensionality reduction compared to other algorithms.