AlgoVision: An LLM-Driven Framework for Visualization and Instructional Explanation for Data Structures and Algorithms Education

Abstract

Data Structures and Algorithms (DSA) represents one of the most demanding courses in undergraduate IT/CS education, yet existing learning tools remain static, non-adaptive, and incapable of generating personalized explanations. This study designed, developed, and evaluated AlgoVision, a web-based LLM-driven framework for generating personalized educational videos. A mixed-methods developmental research design following an Agile iterative methodology was employed across three objectives. A problem validation survey was administered to forty (40) IT/CS students from CCSE at Lorma Colleges. System development integrated Google Gemini 3.1 Flash Lite, Manim, and Kokoro TTS. System acceptance was evaluated using the Technology Acceptance Model, System Usability Scale, and Content Quality instrument among thirty-eight (38) IT/CS students and two (2) DSA instructors from CCSE. The survey confirmed moderate DSA learning difficulties at M = 3.38, Moderate-High Need for alternative explanations at M = 3.93, and Moderate-High Interest in AI-powered solutions from M = 4.11 to M = 4.32. System development yielded a 92.0% video generation success rate and a content quality geometric mean of 0.942. System acceptance demonstrated Moderate-High Acceptance in Perceived Usefulness at M = 6.16, Perceived Ease of Use at M = 6.12, and Behavioral Intention at M = 6.24. The System Usability Scale yielded M = 74.88, rated Excellent, Grade C, and Acceptable. Content Quality averaged M = 4.40. AlgoVision is technically feasible, pedagogically grounded, and highly accepted as a supplementary tool. Future research should address longitudinal learning outcomes and prompt engineering for recursive algorithms.