SIRIEMA: A Framework to Enhance Clustering Stability by Fusing Multimodal Data

Anonymous

TL;DR: This study introduces a new multimodal framework, SIRIEMA, that improves clustering stability in marketing by fusing categorical, numerical, and textual data, utilizing advanced techniques like transformers and generative models

Abstract:

In marketing, customer segmentation is critical for creating content tailored to specific consumer groups. The stability of these segments, hinging on an algorithm's ability to form similar groupings consistently, is essential for effective marketing strategies and higher conversion rates. Traditionally, segment stability can be improved by relying on structured data like age and purchase history and integrating this data with textual information, such as social mnedia posts and product reviews.
This study presents SIRIEMA, a multimodal framework deSIgned to enhance clusteRIng stability by fusing catEgorical, nuMericAl, and textual data.
Our proposal utilizes a transformer-based model for text, data fusion techniques, and generative models like variational autoencoders and generative adversarial networks. Using real-world datasets, SIRIEMA showed enhanced clustering stability and quality compared to existing methods. This research represents a novel approach to customer segmentation and paves the way for future exploration of data fusion techniques in the context of marketing and other applications.

Paper Type: long

Research Area: NLP Applications

Contribution Types: Publicly available software and/or pre-trained models, Data analysis

Languages Studied: English