HI
I have trained a dataset and found that in the training the RMSE=0.078 and in the test RMSE=0.066 ( i used k-cross validation in the training with 5 foldsà
doesn’t mean that there is a problem
because in the course we saw that if the training and the test are high it means that the model is underfitting , and if the training is low and the test is big it is overffiting
but what if the training is high than the test?
and in my case doesn’t mean it’s ok or i should modify my model
Thanks
It could mean your dataset is too small to avoid stastical anomalies.
Can you give a summary of your dataset?
if the human error is small and training error and test error are also small but the test error is higher than the training error by small values it mean that the algorithm is very effective and it is very good
if the human error is small and training error and test error are high but the test error is higher than the training error by small values it mean that the algorithm Suffers from underfit
if the human error is small and training error is also small but test error is higher than the train by high values it mean that the algorithm Suffers from overfit
if the human error is high and training error and test error are also higher than human error by small values it mean that the algorithm is effective and it is good
I hope this photos also help you
please feel free to ask any questions,
Thanks,
Abdelrahman
