Related Experiment Video
Updated: Nov 27, 2025

A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
On Using Linear Diophantine Equations for in-Parallel Hiding of Decision Tree Rules
Georgios Feretzakis1, Dimitris Kalles1, Vassilios S Verykios1
1School of Science and Technology, Hellenic Open University, Patras 263 35, Greece.
This study introduces a novel record augmentation method to protect sensitive patterns in decision trees during data sharing. The technique uses linear Diophantine equations to preserve data utility while hiding critical classification rules.
Area of Science:
- Computer Science
- Data Privacy
- Machine Learning
Background:
- Data sharing is prevalent across industries like marketing and finance.
- Organizations need to protect sensitive patterns within shared datasets.
- Existing privacy methods can reduce data usability.
Purpose of the Study:
- To develop a privacy-preserving method for decision tree induction.
- To hide sensitive classification rules in binary datasets without compromising data utility.
- To offer an alternative to output perturbation and cryptographic techniques.
Main Methods:
- A record augmentation approach is employed.
- A look-ahead technique utilizing linear Diophantine equations is proposed.
- The method adds instances to maintain node entropy.
Main Results:
- The proposed method effectively hides decision tree rules.
- It preserves the initial entropy of nodes, ensuring data usability.
- It offers an optimal solution for hiding one or more rules.
Conclusions:
- Record augmentation with linear Diophantine equations is an effective privacy-preserving technique for decision trees.
- This method balances data privacy with data usability for shared datasets.
- It provides a valuable tool for organizations engaging in data sharing.
Related Concept Videos
Application of Nonlinear Inequalities
Systems of Linear Equations in Two Variables
Statically Indeterminate Problem Solving
Woodward–Hoffmann Selection Rules and Microscopic Reversibility
Synthetic Disvision of Polynomials
Castigliano's Theorem: Problem Solving