CODEWITHSUNDEEP
rmachine-learningsvmlibsvm
user2991421
user2991421

Reputation: 417

How to decrease execution time for SVM

I am currently working on a dataset with 20 features and 400K rows. I'm using e1071 library for multi class classification using SVM. I have this code and it is taking about 2 days to get me the results. Is there a way I can reduce the time complexity using the same library? If not what other library or options should I consider preferably in R?

svm.model <- svm(y ~., data = traindata, gamma = 0.01, cost = 10, kernel= "radial")

Upvotes: 0

Views: 940

Answers (1)

Luis Leal
Luis Leal

Reputation: 3514

I'd suggest try dimensionality reduction(with PCA for example) to reduce the number of features, that would give a performance boost.

Upvotes: 1

Related Questions