Scholarlink

By online processing of streaming data, Online Streaming Feature Selection (OSFS) provides an idea to address critical challenges in big data analysis. In real applications, missing data frequently occurs in streaming data, which will lead to the deviation of feature spatial information. How to establish the relationship between missing data processing and OSFS is the key of Online Sparse Streaming Feature Selection ( \(OS^2FS\) ). To fill this gap, this paper proposes an Online Sparse Streaming Feature Selection Based on Online Correlation Analysis Incorporated Latent Factor Analysis, termed \(OS^2FS_{CALF}\) . The method focuses on the correlation between the feature and the label, and takes this as the basis of missing data processing, and enhances the relationship between missing data processing and feature selection. \(OS^2FS_{CALF}\) uses an \(L_1\) -and- \(L_2\) -norm-oriented LF ( \(L^3F\) ) model that considers label information to process missing data and then conduct feature selection. Experiments in six real-world datasets have shown that can improve the performance of OSFS algorithm on sparse data.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive Correlation-Incorporated Latent Feature Analysis for Online Sparse Streaming Feature Selection

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hao Tian,
  • Yangyi Xia,
  • Jia Wu,
  • Di Wu

摘要

By online processing of streaming data, Online Streaming Feature Selection (OSFS) provides an idea to address critical challenges in big data analysis. In real applications, missing data frequently occurs in streaming data, which will lead to the deviation of feature spatial information. How to establish the relationship between missing data processing and OSFS is the key of Online Sparse Streaming Feature Selection ( \(OS^2FS\) ). To fill this gap, this paper proposes an Online Sparse Streaming Feature Selection Based on Online Correlation Analysis Incorporated Latent Factor Analysis, termed \(OS^2FS_{CALF}\) . The method focuses on the correlation between the feature and the label, and takes this as the basis of missing data processing, and enhances the relationship between missing data processing and feature selection. \(OS^2FS_{CALF}\) uses an \(L_1\) -and- \(L_2\) -norm-oriented LF ( \(L^3F\) ) model that considers label information to process missing data and then conduct feature selection. Experiments in six real-world datasets have shown that can improve the performance of OSFS algorithm on sparse data.