A Classification Algorithm Based on Improved Locally Linear Embedding

Abstract

The current classification is difficult to overcome the high-dimension classification problems. So, we will design the decreasing dimension method. Locally linear embedding is that the local optimum gradually approaches the global optimum, especially the complicated manifold learning problem used in big data dimensionality reduction needs to find an optimization method to adjust k-nearest neighbors and extract dimensionality. Therefore, we intend to use orthogonal mapping to find the optimization closest neighbors k, and the design is based on the Lebesgue measure constraint processing technology particle swarm locally linear embedding to improve the calculation accuracy of popular learning algorithms. So, we propose classification algorithm based on improved locally linear embedding. The experiment results show that the performance of proposed classification algorithm is best compared with the other algorithm.