Skip to content

Recommender engine using SageMaker KMeans clustering for a CloudGuru challenge.

Notifications You must be signed in to change notification settings

msaintfelix/SageMaker_Recommender_Engine_

Repository files navigation

Machine_Learning_SageMaker_Recommender_Engine_

Objective

The objective of this project is to build a recommender engine with AWS SageMaker. The main goal is to recommend the top 20 movie selection to users that are clustered by preference similarities.

For a detailed description of this project, check out my article published by Towards Data Science: https://towardsdatascience.com/machine-learning-recommender-engine-with-aws-sagemaker-4892a9e4a858

Environment and tools

Our MovieLens dataset was downloaded from grouplens.org and more specifically we used movies.csv and ratings.csv from the ml-latest-small.zip file and uploaded it to our AWS S3 bucket. We are working with a Jupyter Notebook with a conda_python3 framework in an AWS SageMaker environment. We used popular ML libraries such as Pandas, Matplotlib and Scikit-Learn to process the data.

Project steps

1. Inspect and visualize the data

In this section, we load the input data into dataframes to gain knowledge of it: dimensions and sizes, outliers.

2. Prepare and transform the data

The next step is to put the data in a format a machine can learn from by combining disjointed data files into one, removing null values, doing some feature engineering.

3. Train

After transforming the data, we can start the training process using the SageMaker built-in K-Means algorithm. The algorithm clusters users so that we can make recommendations.

6. Recommend

Now that we have identified our clusters, we can make recommendations to any user based on weighted ratings from within each cluster.

About

Recommender engine using SageMaker KMeans clustering for a CloudGuru challenge.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published