Figure 3: Various performances of PCA-based unsupervised FE applied to synthetic data. | Scientific Reports

Figure 3: Various performances of PCA-based unsupervised FE applied to synthetic data.

From: Principal Components Analysis Based Unsupervised Feature Extraction Applied to Gene Expression Analysis of Blood from Dengue Haemorrhagic Fever Patients

Figure 3

Upper row (the first synthetic data set, Gaussian): TP, FP, and F-measure for PCA-based unsupervised FE (open black circles), SAM correctly assuming two classes (open red triangles), and SAM wrongly assuming four classes (green crosses).Lower row (the second synthetic data set, compiled from real gene expression data): TP, FP, and F-measures for PCA-based unsupervised FE (open black circles), SAM correctly assuming two classes (open red triangles), and Limma (green crosses). Error bars (95% confidence interval) were less than the size of the characters.

Back to article page