search

gboris / blkbox

Data exploration with multiple machine learning algorithms
14 stars 4 forks source link

feature selection heatmap plotting would be good to select to the desired fs method #48

Closed gboris closed 8 years ago

gboris commented 8 years ago

str(feature_benchmarks_wide$IIZ.0.1) 'data.frame': 6156 obs. of 7 variables: $ dataset : chr "IIZ" "IIZ" "IIZ" "IIZ" ... $ cutoff : chr "0.1" "0.1" "0.1" "0.1" ... $ feature : chr "gene1" "gene10" "gene100" "gene1000" ... $ bartmachine: num 0.0179 0.0131 0.0182 0.0117 0.0166 ... $ GLM : num 0 0 0 0 0 0 0 0 0 0 ... $ PamR : num 0.00988 0 0 0 0 ... $ SVM : num 0 0 0 0 0 0 0 0 0 0 ...

changing the number in the plot funciton solves it. ... [,5:ncol(f....

pdf(paste0("heatmaps_glm", ".pdf"), width = 7, height = 5) for(i in 1:length(feature_benchmarks_wide)){ pheatmap(t(as.matrix(feature_benchmarks_wide[[i]][,5:ncol(feature_benchmarks_wide[[i]])])), main = paste(unique(feature_benchmarks_wide[[i]]$dataset), unique(feature_benchmarks_wide[[i]]$cutoff)), treeheight_col = 0, legend = F, fontsize_col = 0.1, color = brewer.pal(9,"Reds")) } dev.off()

gboris commented 8 years ago

needs a filter >0 since it otherwise utilizes all features which have a value in col 4.

zacdav commented 8 years ago

Each algorithm on the heatmap is a FS method, we arent using it to gauge what to use. This is comparing differences after everything is completed.