Underfitting vs. Overfitting (Simplified ??)

In Machine Learning and in Statistics, there is a concept called "model fit" or "goodness-of-fit" which refers to how closely a model’s predicted values match the observed (or actual) values.

It's like trying on new pants that perfectly fit your waistline. Some might be too loose and others could be too tight. As you try out different pants, you'll encounter some that are way under your current size or other ones that are overly tight for you. This process resembles the concept of model fit in Machine Learning.

A good fit is that sweet spot between underfitting and overfitting. This is a model with only a small number of False Positives and False Negatives. So if the model predicted that something is right, then the actual results are correct (True Positive) and if something is predicted to be wrong, the outputs are also incorrect (True Negative). It allows you to effectively, and precisely predict an outcome, no matter how extensive the data noise and variance are.

Overfitting is an event when a machine learning model learns and takes into account excessive data than necessary. It includes data noise and other variables in your training data to the extent that it negatively impacts the performance of your model in processing new data. There is such an overflow of irrelevant data that affects the actual training data set.

On the other hand, underfitting happens when your model provides poor predictions using your current training data set as well your new data. It fails to effectively generalize your data as it should. This is not a suitable model and causes poor performance with low precision in providing data predictions.

Wanna learn more? You can check out this FREE AWS Machine Learning course that covers the basics of concepts of artificial intelligence and other ML services in AWS!

Grab it here: https://portal.tutorialsdojo.com/courses/aws-machine-learning-services-overview/

You can also check out our collection of free AWS practice tests for the Data Analytics, Machine Learning, Security, and other certification exams.

https://portal.tutorialsdojo.com/product-category/free-practice-exams/

We also have a premium version of our AWS Certified Machine Learning Specialty (MLS-C01) practice tests, with more practice scenarios and detailed explanations for each and every option. It also has bonus flashcards to help you visualize those challenging ML concepts.

Try it out here:

https://portal.tutorialsdojo.com/courses/aws-certified-machine-learning-specialty-practice-exams/