Data, Features and Labels

Every sentence in the rest of this unit uses these words. They are short, they sound alike, and the paper has already caught students out by swapping two of them round. Ten minutes here saves marks in four lessons’ time.

1Data, features, labels

Data

Data is information in any form. A table about fruit is data. Each row holds the information about one fruit, and each fruit is described by certain characteristics.

Features

The columns of the table are the features. In a fruit dataset the features might be colour, size and shape.

Labels

Labelling is attaching meaning to data. One column is special: it is the answer we want the machine to produce. That column is the label.

Which column is the label depends on the problem. If you are trying to work out what fruit this is from its colour, then colour is a feature and the fruit’s name is the label. Ask a different question of the same table and a different column becomes the label.

One table, three ways of looking at it

The first three columns are the features. The last one is the label — the answer we want the machine to give.

Colourfeature	Sizefeature	Shapefeature	Fruitlabel
Red	Small	Round	Strawberry
Red	Medium	Round	Apple
Yellow	Medium	Long	Banana
Green	Medium	Round	Apple
Orange	Medium	Round	Orange
Yellow	Long	Long	Banana

Labelled data. Every row carries its answer. A machine can learn the link between the features and the label from this — which is what supervised learning needs.

2Labelled and unlabelled data

	Labelled data	Unlabelled data
What it is	Data with a tag or label attached	The raw form of the data, with no tag attached
Example	Photos already tagged with who is in them	A folder of photos nobody has named
Used by	Supervised learning	Unsupervised learning

3Training set and testing set

The data you have does not all get used the same way. It is split in two.

Training data set

The collection of examples given to the model to analyse and learn from. It is a set of labelled data — the model can see both the features and the answers.

Testing data set

Used to test the accuracy of the model. The test is performed without the labels showing; the model’s answers are then checked against the real labels.

Teaching, then testing

A teacher teaches a topic through lots of solved examples — that is the training set, questions with answers. Then she sets a class test: questions with the answers withheld, marked afterwards against the answer key. That is the testing set. No teacher sets a test made only of the examples she solved in class, and for the same reason no model is tested on its own training data.

The swap the paper actually used

Training is for learning. Testing is for checking. The 2025–26 sample paper printed the two definitions the wrong way round and asked whether they were correct. Read any statement about these two slowly — the words are nearly identical and the meanings are opposite.

4From the board papers

Statement 1: The testing data set is a collection of examples given to the model to analyze and learn.
Statement 2: The training data set is used to test the accuracy of the model.
Which of the following is correct?CBSE sample paper 2025–26

1 mark

▶Show the answer

Both Statement 1 and Statement 2 are incorrect. They have been swapped. The training set is what the model learns from; the testing set is what its accuracy is measured on.

In supervised learning, what is the purpose of the testing dataset?CBSE board exam 2026

1 mark

▶Show the answer

To evaluate the model’s accuracy. It is not used to train the model and it is not used to create new features — its whole job is to check how well the trained model does on data it has not seen.

5Check yourself

Quick Check

What are the columns of a dataset called?

Quick Check

Data to which no tag is attached is called:

Quick Check

Which set does the model actually learn from?

← Previous

🪆 AI, ML and DL

Rule Based and Learning Based 🔀