A Density-Weighted Information Gain Tree for Clustering Mixed-Type Data

Clustering mixed-type data, which includes both continuous and categorical features, presents significant challenges due to the distinct nature of these data types. Many traditional distance-based and density-based methods struggle with mixed-type data because they are not designed to handle continuous and categorical features simultaneously. To address these limitations, we propose the Density-Weighted Information Gain (DWIG) Tree algorithm, which effectively manages mixed datasets by integrating continuous and categorical features through a recursive partitioning strategy. The DWIG Tree maximizes information gain while accounting for local density variations, resulting in more accurate and interpretable clustering outcomes. Experiments on both synthetic and real-world datasets demonstrate that the DWIG Tree outperforms K-Prototypes, highlighting its superior capability to handle mixed-type data and capture natural groupings more accurately.