Plotting comparison of feature selection

Schrick-Noah_Homework-6.R

@@ -66,6 +66,36 @@ glm.df.1 <- data.frame(att=c("intercept", colnames(bundled_data$train.X)),
 glm.df.1.res <- dplyr::slice_max(glm.df.1,order_by=abs_scores,n=20)
 glm.1.table <- as.data.table(glm.df.1.res)
 
+### Plot
+#### Convert names to indices
+lasso.df$att <- match(lasso.df$att,colnames(bundled_data$train))
+ridge.df$att <- match(ridge.df$att,colnames(bundled_data$train))
+rf.df$att <- match(rf.df$att,colnames(bundled_data$train))
+glm.df.0$att <- match(glm.df.0$att,colnames(bundled_data$train))
+glm.df.1$att <- match(glm.df.1$att,colnames(bundled_data$train))
+
+#### Scale
+lasso.df$abs_scores <- scale(lasso.df$abs_scores)
+ridge.df$abs_scores <- scale(ridge.df$abs_scores)
+rf.df$scores <- scale(rf.df$scores)
+glm.df.0$abs_scores <- scale(glm.df.0$abs_scores)
+glm.df.1$abs_scores <- scale(glm.df.1$abs_scores)
+
+plot(x=lasso.df$att, y=lasso.df$abs_scores, type="l", xlab="Vars", 
+     ylab="Coefficients (Abs Scores)", xaxt="n", col="blue", ylim=c(-1,3),
+     main="Scaled scores for simulated data feature selection")
+axis(1, at=1:101, labels=colnames(bundled_data$train), cex.axis=0.5)
+lines(x=ridge.df$att, y=ridge.df$abs_scores, col="red")
+lines(x=rf.df$att, y=rf.df$scores, col="green")
+lines(x=glm.df.0$att, y=glm.df.0$abs_scores, col="bisque4")
+lines(x=glm.df.1$att, y=glm.df.1$abs_scores, col="purple")
+legend(x="topleft", 
+       legend=c("LASSO", "Ridge", "Random Forest","glmnet (alpha=0)", "glmnet (alpha=1)"),
+       lty=c(1,1,1,1,1),
+       col=c("blue", "red", "green", "bisque4", "purple"),
+       cex=1)
+
+
 ## b. Repeat comparison using a graph with clusters
 if (!require("igraph")) install.packages("igraph")
 library(igraph)