Natural Language Processing and Behavioral Analysis for User-Centric Requirements in Data Warehouse Persona Development

Abstract

The development of effective data warehouse systems remains challenged by inadequate understanding of diverse user requirements, often resulting in low adoption rates and suboptimal return on investment. Traditional approaches to requirements engineering frequently rely on static personas and limited stakeholder engagement, failing to capture the complex behavioral patterns and evolving needs of actual users. This research proposes a novel framework that integrates Natural Language Processing (NLP) and behavioral analysis to create dynamic, data-driven personas for user-centric requirements elicitation in data warehouse environments. The methodology processes multi-source data including query logs, user feedback, and support tickets through an integrated pipeline incorporating topic modeling, sentiment analysis, and ensemble clustering techniques. Experimental validation involving 512 users over a three-month period demonstrates that the framework successfully identifies five distinct user segments with 91.5% accuracy, characterized by unique behavioral patterns and requirement profiles. Requirements developed using data-driven personas show significant improvements in completeness (92.3% vs 73.8%), accuracy (88.7% vs 71.2%), and specificity (94.1% vs 76.5%) compared to traditional methods. The approach reduces requirements elicitation time by 42% and decreases revision cycles by 67%, while achieving a System Usability Scale score of 85.4 versus 68.2 for conventional approaches. The research contributes both theoretical advancements in persona development methodologies and practical solutions for data warehouse requirements engineering. Results confirm that integrating NLP with behavioral analysis enables more accurate user understanding and proactive adaptation to evolving needs.