Proctor Prediction Challenge

The main goal of the challenge is to bridge the gap between geotechnical practice and modern data science methods. Participants learn how to use domain-specific laboratory data (such as grain size distributions and consistency limits) to algorithmically predict soil mechanics parameters—specifically the maximum dry density and optimum moisture content in the Proctor test. The primary objective is to combine sound domain knowledge with statistical modeling to develop robust and generalizable models. 

The competition is open to anyone interested without restriction. It is aimed equally at students, researchers, and practitioners from geotechnical engineering, civil engineering, computer science, and data science. Since mutual learning is the primary focus, participation is open to both individuals and teams of up to five people. 

Registration is free of charge via the kaggle.com platform, where you will also find the dataset.

The data are based on quality-assured laboratory tests from the geotechnical research laboratory at HTWK Leipzig. The provided dataset is approved exclusively for educational and research purposes and non-commercial use. 

Primarily, this competition focuses on skill development: learning how to handle domain-specific data, understanding machine learning methods in the context of geotechnical engineering, and lifting the proverbial veil on "Artificial Intelligence" in civil engineering. In addition, the following awards will be presented to the top three models on the final private leaderboard (Top 3):

• Conference Attendance: Free on-site participation in the 2nd Leipzig Geotechnical Symposium (LeiGS 2026) in Leipzig.
• Publication & Presentation: Publication of your methodology as a short scientific paper in the open-access conference proceedings (including DOI registration) and the opportunity to present the results as a poster at the symposium.

• Competition Start: May 18, 2026
• Submission Deadline: August 16, 2026 (12:00 PM)
• Submission of Conference Papers (Top 3): September 30, 2026
• LeiGS 2026 Conference: November 12 & 13, 2026

Participation is possible at any time.

No. The competition provides an ideal platform to learn how to handle real-world domain data in a practical way or to deepen existing skills. To make getting started as easy as possible, a ready-to-use "Getting Started" notebook in Python is provided. This guides you step-by-step through the entire process, from data import and feature engineering to the creation of an initial baseline model.