In fact, the model I built is a regression model rather than a classification model.
The main issue is the accuracy of training set is far lower than that of testing set.
As is shown in the figures below.
the first figure is the result of training set(the red line is true data and the green is the predictive value)
the second is the result of testing set(the blue line is true data and the green is the predictive value)
the RMSE of traiing set:39.4701
the RMSE of testing set:6.1752
without more information it’s hard to decipher where the issue lies, a few thoughts of what may be causing this:
- training set may feature noisy samples that make convergence in training a tough task
- test set may feature samples that are not representative of the noise in the training set
Is the model showing progress in regards to the loss and metric you are using while training?