Sign inGet started
← Back to all guides

How to do machine learning in a notebook

By Nick Barth

Updated on March 6, 2024

Notebooks have become an indispensable tool for data scientists and AI enthusiasts looking to apply machine learning (ML) techniques. Notebooks provide an interactive environment where one can write code, run it, see the results, and add explanations and visualizations. Below are steps and tips on how to utilize notebooks for your ML projects.

Choosing a notebook platform

There are multiple platforms to choose from when working with notebooks. Jupyter Notebook is among the most popular ones due to its ease of use and flexibility. However, platforms like Google Colab or Deepnote offer cloud-based alternatives that require no installation and come with powerful collaboration features.

Setting up a machine learning environment

Before diving into ML, ensure that you have a proper environment set up. You can install Anaconda, which bundles Jupyter and most of the essential ML libraries like pandas, numpy, scikit-learn, tensorflow, and matplotlib. Alternatively, cloud-based platforms come pre-equipped with these tools, reducing the setup time significantly.

Data preprocessing

The quality of your data is crucial for machine learning. Using notebooks, you can gradually build up your data preprocessing pipeline, which typically includes:

  • Loading data
  • Handling missing values
  • Encoding categorical data
  • Feature scaling
  • Splitting the data into training and test sets

Each step should be contained in a separate cell, allowing for easy modifications and rerunning.

Exploratory data analysis (EDA)

Use the notebook's capability of embedding plots to visualize your data. Through EDA, you'll gain insights that will help you select appropriate features and ML techniques.

Machine learning models


When dealing with classification problems, start by building simple models like a logistic regression or a decision tree to acquire a baseline performance. Gradually transition to more complex models or ensemble methods to improve your predictions and analyze their performance using confusion matrices and ROC curves, all within your notebook.


For regression tasks, begin with linear regression models and progress towards polynomial or support vector regressions as needed. Always evaluate your model performance using appropriate metrics, such as the mean squared error (MSE) or R-squared.

Neural networks and deep learning

For tasks requiring neural networks, Keras and TensorFlow provide easy-to-use APIs. Start by defining your neural network architecture—number of layers and neurons, activation functions, etc. Training a neural network can be visualized in real-time directly in notebooks, which helps in monitoring the training process and making on-the-fly adjustments.

Iterative improvement

One of the key benefits of notebooks is the ease of iteration. Refine your models and preprocessing steps as many times as needed. Leverage version control systems like git to track changes.

Documentation and sharing

An often understated advantage of using notebooks is the ability to interleave code with markdown text and images, making it as much a documentation tool as a coding environment. Make sure to document your analysis and findings, as this will help others understand your workflow and reasoning.


Once your model is ready, you might need to move it from the notebook to a more production-oriented environment. Nonetheless, you can also demonstrate small-scale deployment using web-based applications like Flask or Deepnote apps directly within your notebook, which can provide a proof-of-concept.


Notebooks provide a comprehensive platform for the entire machine learning workflow, from data preprocessing and exploration to model training, evaluation, documentation, and deployment. Whether you're working on classification, regression, or neural networks, notebooks allow for a flexible, iterative, and accessible way to refine your ML models while ensuring your work is understandable and sharable.

That’s it, time to try Deepnote

Get started – it’s free
Book a demo



  • Integrations
  • Pricing
  • Documentation
  • Changelog
  • Security




  • Privacy
  • Terms

© Deepnote