Adaptive Neighborhoods Locality Preserving Projections

In contemporary society, the analysis and utilisation of big data is of paramount importance. However, traditional feature extraction methods frequently encounter the curse of dimensionality when dealing with high-dimensional data, which is no longer able to adapt to the current needs of practical applications. The manifold learning method, which is capable of effectively extracting features of high-dimensional data, ne-cessitates the resolution of the problem of how to choose an appropriate local neighbourhood. Consequently, this paper proposes an Adaptive neighborhood Locality Preserving Projection (LAPP) based on the Locality Preserving Projection (LPP). TThis method combines two neighborhood selection method. To achieve an adaptive radius selection for local neighborhoods of sample points based on the distance characteristics between sample points. The LAPP proposed in this paper markedly reduces the dimensionality reduction and enhances the computational efficiency while maintaining a certain degree of accuracy.