An Expert-Based Approach for Recommending Data Visualization Techniques
Date Issued
2025-04
Author(s)
Gafarov, Kanan
Abstract
In the era of data-driven decision-making, effective visualization of complex datasets plays a
pivotal role in enhancing analytical insights and communication. Yet, selecting the most appropriate
data visualization technique remains a significant challenge for users, especially those without formal
training in data analytics or visual design. This research addresses that gap by developing a rule-based
expert recommendation system that guides users in selecting suitable visualization techniques based on
specific dataset characteristics. The proposed system integrates expert knowledge, formalized rules, and
a user-centric interface to support and streamline the visualization selection process.
The system architecture follows a Model–View–Controller (MVC) design pattern, with a robust
backend built using Python and MongoDB. The Model stores a curated database of visualization
techniques along with associated rules and attributes, the Controller implements a dynamic decision
engine that interprets user responses, and the View presents an interactive Q&A interface built using
Streamlit. The questionnaire comprises 8–10 targeted questions that gather key information about the
dataset—such as data type, structure, analytical goal, and data volume—which are then evaluated by
the system’s reasoning engine to generate tailored recommendations.
A unique feature of this system is its expert rule base, manually constructed based on visualization
theory, best practices, and authoritative resources. The database of visualization techniques was also
built from scratch, requiring extensive research and careful structuring of technique-specific metadata.
The rules were refined through iterative testing and feedback to ensure accuracy and contextual
relevance across a variety of datasets.
The system was evaluated through a formative usability study with five participants, who tested
the prototype on real datasets and provided qualitative feedback. Results indicated high usability, with
participants praising the intuitive question flow, clarity of recommendations, and educational value.
Users found that the system either confirmed their expectations or introduced new and contextually
relevant visualization ideas. The rule-based recommendations were transparent and interpretable,
contributing to user confidence and improved visualization literacy.
This research contributes a novel and practical solution to the problem of visualization technique
selection by embedding expert knowledge in a usable and accessible form. It has implications for both
novice users seeking guidance and experienced analysts looking for validation or inspiration. Future
enhancements may include expanding the visualization database, improving rule transparency, and
deploying the system as a web-based application. Overall, the system demonstrates the feasibility and
value of expert-based, rule-driven approaches in supporting effective data visualization.
pivotal role in enhancing analytical insights and communication. Yet, selecting the most appropriate
data visualization technique remains a significant challenge for users, especially those without formal
training in data analytics or visual design. This research addresses that gap by developing a rule-based
expert recommendation system that guides users in selecting suitable visualization techniques based on
specific dataset characteristics. The proposed system integrates expert knowledge, formalized rules, and
a user-centric interface to support and streamline the visualization selection process.
The system architecture follows a Model–View–Controller (MVC) design pattern, with a robust
backend built using Python and MongoDB. The Model stores a curated database of visualization
techniques along with associated rules and attributes, the Controller implements a dynamic decision
engine that interprets user responses, and the View presents an interactive Q&A interface built using
Streamlit. The questionnaire comprises 8–10 targeted questions that gather key information about the
dataset—such as data type, structure, analytical goal, and data volume—which are then evaluated by
the system’s reasoning engine to generate tailored recommendations.
A unique feature of this system is its expert rule base, manually constructed based on visualization
theory, best practices, and authoritative resources. The database of visualization techniques was also
built from scratch, requiring extensive research and careful structuring of technique-specific metadata.
The rules were refined through iterative testing and feedback to ensure accuracy and contextual
relevance across a variety of datasets.
The system was evaluated through a formative usability study with five participants, who tested
the prototype on real datasets and provided qualitative feedback. Results indicated high usability, with
participants praising the intuitive question flow, clarity of recommendations, and educational value.
Users found that the system either confirmed their expectations or introduced new and contextually
relevant visualization ideas. The rule-based recommendations were transparent and interpretable,
contributing to user confidence and improved visualization literacy.
This research contributes a novel and practical solution to the problem of visualization technique
selection by embedding expert knowledge in a usable and accessible form. It has implications for both
novice users seeking guidance and experienced analysts looking for validation or inspiration. Future
enhancements may include expanding the visualization database, improving rule transparency, and
deploying the system as a web-based application. Overall, the system demonstrates the feasibility and
value of expert-based, rule-driven approaches in supporting effective data visualization.