Home → Articles → How to Create a Product Recommendation System with Python, Pandas, NumPy, and Scikit-learn

How to Create a Product Recommendation System with Python, Pandas, NumPy, and Scikit-learn

20 Mar, 2026

Introduction

Recommendation systems help businesses suggest products to customers based on their preferences and behavior. With Python libraries like Pandas, NumPy, and Scikit-learn, you can build a simple recommendation engine that analyzes user ratings and predicts what products they might like.

This guide shows you how to create a basic product recommendation system step by step.

Prerequisites

Before you start:

• Purchase an Ubuntu 24.04 VPS server. If you don't have a VPS server, sign up with Vultr and get up to $300 worth of free credit to test the Vultr platform.

• SSH to your VPS server using PuTTY for Windows or run the following command if you're using Linux or Mac.

  console

  $ ssh username@vps_server_public_ip_address
  

• Create a non-root user with sudo privileges. Read our guide on How to Create a Non-Root Sudo User on Ubuntu 24.04. You'll use this user's account to run the commands in this guide.

• Install Python 3.10 or later by following our How to Install Python on Ubuntu 24.04 guide.

Set Up a Virtual Environment

It’s best practice to use a virtual environment so your project dependencies don’t interfere with other Python projects.

1. Create and switch to a new directory for your project

   console

   $ mkdir product-recommendation && cd product-recommendation
   

2. Create a virtual environment:

   console

   $ python3 -m venv venv
   

3. Activate the virtual environment:

   console

   $ source venv/bin/activate
   

4. Install the required libraries inside this environment:

   console

   (venv) $ pip install pandas numpy scikit-learn
   

Create a Sample Dataset

1. Use Pandas to build a simple dataset of users rating products.

   Python

   import pandas as pd
   
   # Sample data: user, product, rating
   data = {
       'user_id': [1, 1, 1, 2, 2, 3, 3, 4],
       'product_id': [101, 102, 103, 101, 104, 102, 104, 103],
       'rating': [5, 3, 4, 4, 5, 2, 4, 5]
   }
   
   df = pd.DataFrame(data)
   print(df)
   

2. You'll use this data set in the next step.

Create a User-Product Matrix

1. Pivot the data so that each row represents a user and each column represents a product.

   Python

   user_product_matrix = df.pivot_table(
       index='user_id',
       columns='product_id',
       values='rating'
   )
   
   print(user_product_matrix)
   

This matrix shows which user rated which product. Missing values mean the user hasn’t rated that product yet.

Handle Missing Values

1. Replace missing values with 0 for simplicity.

   Python

   import numpy as np
   
   matrix_filled = user_product_matrix.fillna(0)
   print(matrix_filled)
   

Compute Similarity Between Users

1. Use cosine similarity from Scikit-learn to measure how similar users are based on their ratings.

   Python

   from sklearn.metrics.pairwise import cosine_similarity
   
   # Compute similarity
   user_similarity = cosine_similarity(matrix_filled)
   
   print(user_similarity)
   

2. Each value shows how similar two users are (closer to 1 means more similar).

Make Recommendations

1. Recommend products to a user based on what similar users liked.

   Python

   def recommend_products(user_id, matrix, similarity, top_n=2):
       # Get similarity scores for the user
       sim_scores = similarity[user_id - 1]  # adjust index
       # Find most similar user
       similar_user = np.argsort(sim_scores)[-2] + 1  # skip self
       print(f"User {user_id} is most similar to User {similar_user}")
   
       # Get products rated by the similar user
       user_ratings = matrix.loc[similar_user]
       recommended = user_ratings[user_ratings > 0].index.tolist()
   
       return recommended[:top_n]
   
   print(recommend_products(1, matrix_filled, user_similarity))
   

2. This function finds the most similar user and recommends the products they rated.

Test the Recommendation System

1. Try recommending products for different users:

   Python

   print("Recommendations for User 1:", recommend_products(1, matrix_filled, user_similarity))
   print("Recommendations for User 2:", recommend_products(2, matrix_filled, user_similarity))
   

Conclusion

In this guide, you built a simple product recommendation system using Pandas, NumPy, and Scikit-learn. You learned how to create a user-product matrix, compute similarity between users, and recommend products based on similar users’ ratings. This is a basic collaborative filtering approach. You can improve it by using larger datasets, more advanced similarity measures, or machine learning models to achieve higher accuracy.