Last active
August 29, 2015 14:10
-
-
Save mkowoods/797e8b9bf446fdcd209f to your computer and use it in GitHub Desktop.
Playing with Test and Training Set Error
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
n_col = 100 | |
n_rows = 1000 | |
set.seed(500) | |
rand.matrix <- matrix(data = rep(NA, n_rows*n_col), nrow = n_rows, ncol = n_col) | |
rand.class <- rbinom(n = n_rows, size = 1, prob = 0.7) | |
#Assign some variables to be randomly relevant | |
for(i in 1:n_col){ | |
predictor <- rnorm(n_rows) | |
if(i <= 5){ | |
rand.matrix[, i] <- predictor + rand.class*rnorm(n_rows) | |
}else{ | |
rand.matrix[, i] <- predictor | |
} | |
} | |
rand.data <- as.data.frame(cbind(rand.class, rand.matrix)) | |
training <- sample(1:n_rows, 0.6*n_rows) | |
test <- -sample(1:n_rows, 0.6*n_rows) | |
training.acc <- rep(NA, n_col) | |
test.acc <- rep(NA, n_col) | |
ordered.cor <- order(cor(rand.data)[1:(n_col + 1), 1], decreasing = T) | |
for(i in 1:n_col){ | |
m1 <- glm(rand.class~., rand.data[training, ordered.cor[1:(1+i)]], family="binomial") | |
training.acc[i] <- sum((predict(m1, rand.data[training,], type = "response") > 0.5) == rand.data[training,]$rand.class)/length(training) | |
test.acc[i] <- sum((predict(m1, rand.data[test,], type = "response") > 0.5) == rand.data[test,]$rand.class)/(n_rows - length(training)) | |
} | |
tmp <- data.frame(idx = 1:n_col, training.acc = training.acc, test.acc = test.acc) | |
g <- ggplot(data = tmp, aes(idx)) | |
g <- g + geom_point(aes(y = training.acc)) +geom_smooth(aes(y = training.acc), method = "loess") | |
g <- g + geom_point(aes(y = test.acc), color = "red") +geom_smooth(aes(y = test.acc), method = "loess") | |
g <- g + ylab("accuracy") | |
print(g) | |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment