SERIEMA: A Framework to Enhance Clustering Stability, Compactness, and Separation by Fusing Multimodal Data

This article presents SERIEMA, a multimodal framework deSignEd to enhance clusteRIng stability, compactness, and separation by fusing catEgorical, nuMericAl, and textual data. We aim to overcome these critical challenges in clustering, which is essential for marketers to effectively provide targeted content to various consumer segments. SERIEMA aims to provide a more nuanced and comprehensive understanding of customer segments by leveraging a transformer-based embedding model for textual data analysis, a data fusion component, and a generative-based model. This integration overcomes the limitations of traditional methods that rely solely on structured data or text, facilitating precise segmentation and improved marketing strategies. The framework is evaluated using established stability measures and benchmarked against existing strategies across real-world datasets. The results highlight SERIEMA’s superior effectiveness in enhancing clustering stability, compactness, and separation compared to traditional segmentation methods. The study’s contributions are significant, marking a notable advancement in multimodal learning and customer segmentation.