Why do we split data into train test and validation sets?

To train machine learning models we need to provide the model with a training and testing set. And sometimes even a validation set. These terms tend to be used interchangeably causing confusion. So once and for all, let's learn what each of these data splits do and how they contribute to model development.

👋 Keep in touch?
==========================

Courses & resources
============================
📙 Fundamentals of Deep Learning in 25 pages

👩‍💻 Hands-on Data Science: Complete your first portfolio project

📥 Streamlit template

🤖 Deep Learning 101 with Python and Keras (FREE)

🏃‍♀️ Data Science Kick-starter mini-course (FREE)

🐼 Pandas cheat sheet (FREE)

📝 NNs hyperparameters cheat sheet (FREE)

[iaboodws]

Just what I was looking for, your video is so simple and easy to understand, and straight to the point!!!

[jamalnuman]

Thanks for making a distinction between testing and validation

[sapnilpatel]

video is very much useful. Your channel is so underrated.

[syahwiza]

Thanks for your explanation, it is very useful for me.

[facundostratocaster]

Simple and good explanation, thank you so much ☺️

[SocialAviation]

I love your content. Everytime I split my data into train and valid, either using trainsplit function or manually, my val loss does not decrease below 1. The only way to get my val loss lower and lower, is to use part of my train data as validation data 😢

[jamesadeke]

Good day ma, please can you help me out? I have been trying to figure out this for a long time but i could not. I want to know the best evaluation plots for machine learning models, specifically for classification problems. How best can someone visualize performance? Unlike deep learning models, you can use train and test curves, how best can we visualize using machine learning models? Do you have any video you have done about that? been checking your playlists but i can't find such, kindly help us out. Thanks

[Randomsi_]

This is very helpful on my ongoing thesis🥹

[misha]

After finding the best hyperparameters for a model using validation data, should we retrain the model using both the training and validation data before using it on the test data?

[_suryakantdhote]

please make a video on logestic regression

[jamalnuman]

what is the need of testing data is the hyperparameters don't to be optimized?

[jameshopkins]

Can you explain something about
Example the meaning and useful of each one

[jsingh]

Can you make a project use of some ML application

[bay-bicerdover]

0:50'de blop efekti ödümü kopardı

[SyamKishoreNaidu]

Can we expect more pandas related videos

[babaabba]

Ah, if you were living in France, I would have married you immediately, I would have taken you to some fancy restaurant everyday and during the night, you would have done my assignments within data science.