Fuzzy SLIQ decision tree algorithm.

Area of Science:

• Machine Learning
• Artificial Intelligence
• Data Mining

Background:

• Traditional decision tree algorithms generate sharp decision boundaries, which are often unrealistic for real-world classification tasks.
• Large and complex decision trees can lead to difficult-to-interpret rules, hindering practical application.

Purpose of the Study:

• To propose a novel fuzzy supervised learning in Quest (FS-DT) decision tree algorithm.
• To develop a method for constructing fuzzy decision boundaries instead of crisp ones.
• To reduce the size of decision trees for improved comprehensibility and efficiency.

Main Methods:

• Modification of the existing SLIQ (Supervised Learning in Quest) decision tree algorithm.
• Implementation of fuzzy logic to create decision boundaries.
• Comparative performance analysis using real-life datasets from the UCI Machine Learning Repository.

Main Results:

• The proposed FS-DT algorithm successfully constructs fuzzy decision boundaries.
• FS-DT demonstrates superior classification accuracy compared to the crisp SLIQ algorithm.
• A significant reduction of over 70% in decision tree size was achieved by FS-DT.

Conclusions:

• The FS-DT algorithm offers an effective approach to handling real-world classification problems by employing fuzzy decision boundaries.
• FS-DT provides a more accurate and interpretable alternative to traditional crisp decision trees.
• The substantial reduction in tree size makes FS-DT a more practical and efficient solution.