Data package accompanying the publication "Machine learning short-ranged many-body interactions in colloidal systems using descriptors based on Voronoi cells". It provides all the necessary resources to reproduce the results and analyses presented in the study, which introduces a machine learning (ML) strategy for accurately modeling highly local many-body interactions in colloidal systems. Specifically, for a two-dimensional system consisting of polymers and colloids, we developed a Voronoi-based description of the system and demonstrated that it accurately captures the many-body nature of the system. This data package contains all relevant codes, data and analysis notebooks te reproduce the results found in the paper. The package has the following structure: [FIGURES] contains all figures from the paper, along with the relevant analysis notebooks, Adobe Illustrator files, and additional data to generate the figures. [CODES] contains the scripts to generate the training data, to train the machine learning models, and to run both the reference, brute force system and the machine learning system. [DATA] contains the training data, the pre-trained models, and simulation output from both the brute force system an d the machine learned system. Each directory contains a README.txt file that describes the content of the directory.