0

When i was doing prediction in h2o,i realized that all predict results are totally the same for different days in test.I imported the train and validation data feed the network, and set the nflods number as 10. According to my understanding, the prediction result should be various depends on days. But the results are totally the same. given input and output in validation and train data set are different, do you know why the predictions are the same, any adjustments i should do for more data ? Thnaks very much for your reading and helping !!!

hanting du
  • 11
  • 4
  • 1
    Without seeing your code, workflow, or the input data, it's really hard to help. See https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example – be_green Aug 21 '17 at 15:30
  • Are you looking at the predicted class or prediction itself? If the threshold is bad, then it might score everything as the same class (even though the predicted values will be different, which means you need a better threshold). Sorry, but I can't help w/o seeing reproducible code and what your output looks like. – Erin LeDell Aug 21 '17 at 22:24

0 Answers0