American Express Interview Question | Overfitting
Question
How will you make sure that your model is not undergoing any type of overfitting? State some ways to avoid it.
in progress
0
Machine Learning
55 years
4 Answers
1282 views
Great Grand Master 0
Answers ( 4 )
If the validation set accuracy is less than the training set accuracy then we can know that there is overfitting. To prevent overfitting, we can use regularization L1 or L2. If we are using a neural network we can introduce dropout. We can also use early stopping to prevent overfitting.
We can use cross validation. We can also try to use simple models like logistic regression instead of complex model like random forest.
1) Cross validation
2) Feature Selection
3) Regularization
Common ways of avoiding overfitting:
1. Cross-validation
2. Train with more data
3. Removing features (Unnecessary features)
4. Regularization
5. Ensembling
If the performance of test test is very low compared to training set then we know that Overfitting is there.
To overcome overfitting we can:
1. Use Regularization- Lasso and Ridge
2. Hyper-Parameter tuning
3.Cross Validation
4.Using ensemble Technique
5.If the no. of columns>>> no of rows then reducing dimension can help
6.If the model is sensitive to imbalanced dat like KNN then consider resampling