I would like to predict soil classes with artificial neural network (ANN) in the R environment. I train my model with neuralnet package. When using the predict function, I will only get the probability map for soil classes, not the class-map itself. Even though, in predict function, I specified type="class". So, how can I get the pure classes?
Try your prediction function without any condition for "type". The default output for the prediction function is class. It might help.
Dear Shams,
the returned value of neuralnet::predict.nn() according to its help is "Matrix of predictions. Each column represents one output unit." So you should do something like this:
apply(X = predict(), MARGIN = 1, FUN = function(probabilities) classes[which.is.max(probabilities)])
, where classes is the vector of the class names.
HTH,
Ákos
Perhaps consider assigning numbers to the different classes and then training your model on the assigned numbers. Based on the predictions which will be in number form you can transform back to classes.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package