Train → Poor
Validation → Poor

Likely model signalই ধরতে পারছে না।

Possible solutions:

Better features
More expressive model
Reduce regularization
Add nonlinear features
Train longer