Dimensionality scale back in massive datasets using PDLPP

The main objective of this paper is to reduce data dimensionality in high-dimensional feature datasets. It uses an effective distance based Non-integer Matrix Factorization (NMF) method to resolve the problem of data dimensionality. The non-orthogonality arising due to increasing dimensionality is resolved using NMF and an effective distance measurement. This process involves organizing the datasets to form a defined geometric structure since conventional dimensionality reduction principles capture the structured data using a similarity matrix with a distance-based measurement. However, such distance-based measurements cannot fit dynamic data structure to the model and most of the intrinsic structure of the data is ignored. Hence, to avoid this complexity, the proposed method uses Probabilistic Distance Locality Preserving Projections (PDLPP) to structure the dynamic data. The proposed method is evaluated against the conventional methods in terms of its accuracy and normalized mutual information over different test cases. The proposed method increases the performance of learning the patterns in high dimensional data with less computation time. The results demonstrated that the proposed method fits well with static and dynamic data to query the objects in the search space.