THE IMPACT OF CATEGORICAL DATA ENCODING METHODS ON ARTIFICIAL INTELLIGENCE ALGORITHMS
Authors/Creators
Description
This study analyzes the effectiveness of categorical data encoding methods in artificial intelligence algorithms. The research examines the operational characteristics and impact on results of four widely used encoding techniques—Label Encoding, One-Hot Encoding, Frequency Encoding, and Target Encoding—applied to Decision Tree and K-Nearest Neighbors (KNN) algorithms. Using a real-world dataset, each encoding method was applied separately and evaluated with both algorithms. Model performance was assessed using conventional evaluation metrics such as accuracy, precision, recall, and F1-score. The results indicate that the combination of encoding method and selected algorithm has a significant effect on model quality. In particular, One-Hot Encoding yielded the best results with Decision Trees, while Target Encoding was found to be most effective for the KNN algorithm. The study concludes by outlining important considerations and practical recommendations for selecting appropriate encoding methods.