返回
A correlation analysis framework via joint sample and feature selection
DOI:10.1007/s11042-022-14237-5.png)
摘要
En 中文
Correlation Analysis is a popular technique for describing relationships between two datasets. In this paper, we proposed a correlation analysis framework via Joint Sample and Feature Selection (CAF-JSFS). Different from traditional correlation analysis where only feature selection is considered and each data point is treated equally, the significance of each data point is measured by a sample selection strategy in this framework. Considering that the principal projection is a feasible representation of data, the relationship between this principal projection and each data sample is recursively learnt through two sample selection strategies: cosine similarity and total distance metrics. In addition, CAF-JSFS solves the problem of feature redundancy caused by sample feature selection, and eliminates irrelevant features, thereby improving classification accuracy. This enhances the discriminative power of CAF-JSFS in noisy scenarios which makes better correlation projections achievable to improve performance. Extensive experiments on several datasets demonstrated the effectiveness of the proposed method compared to the state-of-the-art methods.
Keyword:
Correlation analysis
Sample selection
Feature selection
Cosine similarity metric
Total distance metric
期刊
IF:
3
论文数:
1.9W
被引数:
3.2W
机构
引用论文
Isatuximab, carfilzomib, lenalidomide, and dexamethasone (Isa-KRd) in front-line treatment of high-risk multiple myeloma: interim analysis of the GMMG-CONCEPT trialIsatuximab、carfilzomib、lenalidomide和dexamethasone(Isa-KRd)在高危多发性骨髓瘤一线治疗中的应用:GMMG-CONCEPT试验的 interim analysis
Effect of crystal, grain, and particle size on X-ray power diffracted from powders晶体、晶粒和颗粒尺寸对粉末X射线衍射功率的影响
Influence of Menopause and Lifestyle Factors on High Density Lipoproteins in Middle-Aged Women
Menopause
IF0

