CODEWITHSUNDEEP
rmachine-learningclassificationsvmprediction
ltrd
ltrd

Reputation: 326

SVM strange classification

I would like to have a prediction of short term direction of a certain asset. I did an SVM model based on Dane_train data.frame (8058 rows):

SVM1 <- e1071::svm(Direction ~ logReturns_1,
               data = Dane_train,
               type = "C-classification",
               cost = 0.1,
               kernel = "linear",
               scale = FALSE)

And then I would like to make a prediction of the records from Dane_test data.frame:

pred <- predict(SVM1, data = Dane_test, method = "class")

The first strange thing is that I get factor vector of length 8058 - not 2015 from test data. Then when I do:

table(pred, Dane_test$Direction)

I have an error because of the inappriopriate length of the arguments. What have I done wrong?

Upvotes: 2

Views: 50

Answers (1)

G5W
G5W

Reputation: 37661

The arguments for predict are a little different. You need newdata=Dane_test not data=Dane_test.

Upvotes: 2

Related Questions