classification - class importance for random forest in r -

I used randomForest pkg r to predict binary orbit based on 11 numerical predictions I am doing Of the two classes, hits or miss , class hits is of more importance, that is, how often I want to know the prediction code. > Hit

What is the importance of hits to train the random forest? Currently trained random forest only estimates exactly 7% of hits and is definitely like improvement.

High importance? I do not know how to tell any algorithm at this time, "I'm not joking" : I think this analysis should be correct. "

You are always making verses vs. bias battle If you improve training accuracy too much, then you run the risk of overfitting.

You can adjust the random forest by changing the size of the random sample of predictions. If you have meter predictions, then the recommendation for random forest is for the number of division in the pe = M = 1/2 tree. You can also change the number of trees.

You can also try another algorithm, such as (normalized enhanced retrograde model) or

how does it work when you plot it? When you look in Scatterplot, do any explicit groups jump over you?

Despite the algorithm, I recommend that you verify the n-fold of your model.