Visualization of large-scale user association feature data based on a nonlinear dimensionality reduction method

Abstract The prosperity of data science and the booming growth of the internet industry have made the analysis and processing of large-scale user-related feature data a particularly important issue in modern society. Traditional linear dimensionality reduction methods often fail to obtain intuitive information and discover intrinsic connections and patterns in complex high-dimensional data. This study proposes a large-scale user association feature data (LSUAFD) visualization study based on nonlinear dimensionality reduction methods to address the complexity and high-dimensional issues of data. First, a LSUAFD visualization platform based on the nonlinear dimensionality reduction method is constructed. Secondly, the nonlinear dimensionality reduction algorithm is integrated into the platform, and the functional modules and data dimensionality reduction modules of the platform are finally tested. The results indicated that by combining principal component analysis with nonlinear dimensionality reduction algorithms (t-SNE and LargeVis) to provide interactive parameter adjustment and visualization in the front end, the selected dataset exhibited excellent analytical performance. Furthermore, the method could stably and accurately detect outliers within the dataset. The errors were primarily distributed between 0.4 and −0.4, indicating a high level of precision in the interpretation of data distributions.