{"id":177192,"date":"2026-02-18T22:32:04","date_gmt":"2026-02-18T21:32:04","guid":{"rendered":"https:\/\/liora.io\/en\/?p=177192"},"modified":"2026-02-18T22:32:05","modified_gmt":"2026-02-18T21:32:05","slug":"train_test_split-tutorial-on-how-to-use-this-function","status":"publish","type":"post","link":"https:\/\/liora.io\/en\/train_test_split-tutorial-on-how-to-use-this-function","title":{"rendered":"“train_test_split: Tutorial on how to use this function"},"content":{"rendered":"

A Machine Learning model is capable of learning autonomously from one dataset, with the aim of predicting behavior on another dataset. To do this, it finds underlying relationships between independent explanatory variables and a target variable in the initial dataset. It then uses these patterns to predict or classify new data. <\/strong><\/p>\n\n\n\n

How do I define the train_test_split function?<\/h2>\n\n\n\n

To verify the effectiveness of a Machine Learning model,<\/a> the initial dataset is divided into two sets: a training set and a test set. The training set is used to fit, i.e. train, the model on part of the data. The test set is used to evaluate the model’s performance on the other part of the data. The train_test_split function in Python’s ScikitLearn (sklearn) library allows this separation into two sets.<\/a><\/p>\n\n\n\n

First of all, you need to import the train_test_split function from sklearn’s model_selection package using the following code:<\/p>\n\n\n\n

from sklearn.model_selection import train_test_split<\/code><\/pre>\n\n\n\n
Once imported, the function takes several arguments:<\/strong><\/p>\n\n\n\n
\"\"<\/figure>\n\n\n\n
1) Arrays extracted from the dataset to be split.<\/h3>\n\n\n\n
In supervised learning<\/strong>, these arrays are the input array X, consisting of the explanatory variables in columns, and the output array y, consisting of the target variable (i.e. the labels).<\/p>\n\n\n\n
In unsupervised learning, the only array in argument is the input array X, made up of the explanatory variables in columns.<\/p>\n\n\n\n
Note:<\/strong> Be careful with dimensions! X must be a two-dimensional array. y must be a one-dimensional array equal to the number of rows in X. To do this, use the .reshape function.<\/p>\n\n\n\n
2) Test set size (test_size) and training set size (train_size).<\/h3>\n\n\n\n
The size of each set<\/strong> is either a decimal number between 0 and 1 representing a proportion of the dataset<\/a>, or an integer number representing a number of examples in the dataset.<\/p>\n\n\n\n
Note:<\/strong> It is sufficient to define just one of these arguments, the second being complementary.<\/p>\n\n\n\n
3) The random state (random_state).<\/h3>\n\n\n\n
The random state is a number that controls how the pseudo-random generator divides the data.<\/strong><\/p>\n\n\n\n
Note<\/strong>: Choosing an integer as the random state ensures that the data is split in the same way each time the function is called. This makes the code reproducible.<\/p>\n\n\n\n
4) Le shuffle (shuffle).<\/h3>\n\n\n\n
Shuffle is a Boolean that selects whether or not data should be shuffled before being separated. If the data is not shuffled, it is separated according to the order in which it was initially displayed.<\/p>\n\n\n\n
Note:<\/strong> The default value is True.<\/p>\n\n\n\n
5) Stratify (stratify).<\/h3>\n\n\n\n
The Stratify parameter selects whether the data are separated<\/strong> so as to keep the same proportions of observations in each class in the train and test sets as in the initial dataset.<\/a><\/p>\n\n\n\n
Remarks:<\/strong><\/p>\n\n\n\n