Open sagnik1511 opened 2 years ago
I would like to suggest that we can scale the data directly by using Sklearn.preprocessing
scaler=MinMaxScaler() x_train=scaler.fit_transform(x_train) x_val=scaler.transform(x_val)
by adding these lines to the training.py file we can easily scale the data without needing to parse through each feature through a for loop which would be more time consuming and also can be a reason for many bugs. we can also put conditions if the model is for regression and also ask user which scailing function they want and apply those to the x_train and x_val
If you agree to this idea then please assign this Issue to me.
@Tihsrah , it is not better to use minmaxscaling for continuous data , so in some cases it it better to just scale down or scale up. So basically the function should be flexible for every single data column.
If you have a flexible idea about it, please drop the idea in the comments, if it shows clarity, I'll assign you.
What if we first do a "Robust scalar" over the data and then use "Standard Scaler"
@Tihsrah, I would suggest you be flexible while making the class as I stated early.
Follow contributing guidelines on README.md