Scribd
Upload
5 views11 pages

Understanding Overfitting and Underfitting

The document discusses overfitting and underfitting in machine learning models. Overfitting occurs when a model performs well on training data but poorly on new data, often due to noise, while underfitting happens when a model fails to perform adequately on both training and new data due to insufficient or incorrect data. Solutions to overfitting include pruning, cross-validation, and regularization to improve model accuracy and generalization.

Uploaded by

SK.Kalesha
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
5 views11 pages

Understanding Overfitting and Underfitting

The document discusses overfitting and underfitting in machine learning models. Overfitting occurs when a model performs well on training data but poorly on new data, often due to noise, while underfitting happens when a model fails to perform adequately on both training and new data due to insufficient or incorrect data. Solutions to overfitting include pruning, cross-validation, and regularization to improve model accuracy and generalization.

Uploaded by

SK.Kalesha
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Overfitting

Overfitting

• Refers to a model which performs very well w.r.t.


training set. Such model is able to predict the noise
and randomness w.r.t. training set very well
• But would not perform well when it comes to newly
arrived data.
• Occurs with nonparametric and nonlinear models.
Underfitting

• Refers to a model that does not perform well


both with training set and newly arrived data.
• An underfitted model would be directly rejected
due to its performance via benchmark and
accuracy.
Reasons for underfitting occurrences

• Under fitting is said to happen when a model is


unable to capture or understand the nature of
data. This has a direct impact on the accuracy
of the model.
• The reason for this could be:
• Less data available to analyze
• Wrong data
Reasons for underfitting occurrences

• Wrong features considered


• It could be also be due to wrong formed model
• The solution to under fitting is to make use of more
data and possibly even clean data by removing
unwanted features or columns.
How under fitting looks when plotted:
Does over fitting get affected by noise?

• Over fitting is said to happen when a model


gets easily affected by noise or outliers.
• The details available in such a case are more,
and as a result of this, model gets affected.
The over fitting diagram when plotted:
Over fitting can be overcome by:

1. Pruning: Pruning as the name suggests, is


cutting. We prune wanted data in addition to
nodes that gets easily affected by wrong data.
Over fitting can be overcome by:

[Link]-Validation: Sample prediction error is one way


that helps on the problem of over fitting.
This is generally accomplished with the help of k fold
validation. In k fold validation, the original sample data is
categorized into k subsets.
One of the samples is used for testing, and remaining
subsets are used to form the model. The output results
which would be collected and then averaged out get the
final estimation.
Over fitting can be overcome by:

3. Regularization: The aim of this is to find


out features that align with the objective
of the problem and thus removing
features that do not contribute to the final
output.

You might also like