Learning Path Recommendation Method Based on Improved Particle Swarm Optimization with Elite Learning and Adaptive Mutation

Abstract

To address the problem that personalized learning path recommendation in online learning struggles to balance knowledge logic constraints with learner adaptation, this paper proposes a personalized learning path recommendation method based on an improved particle swarm optimization algorithm (IHPSO). First, taking the "C Programming" course as an example, this method constructs a course knowledge graph containing 197 knowledge points and 189 prerequisite relations, and generates a knowledge point sequence conforming to instructional logic through topological sorting. Second, a learner model is constructed from three dimensions: cognitive level, learning style, and available learning time, and a comprehensive objective function is designed to quantify the degree of path adaptation. Finally, three improvements are made to the standard particle swarm optimization algorithm: introducing an elite learning strategy to enhance convergence performance, designing an adaptive mutation mechanism to maintain population diversity, and proposing heuristic local search to achieve fine-grained path optimization. Experimental results show that IHPSO achieves an average fitness of 0.2039, which is 62.46% lower than that of the standard binary particle swarm optimization algorithm (BPSO) and 20.17% lower than that of the genetic algorithm (GA), while the convergence speed is improved by 65% compared with GA. The differences in all metrics are statistically significant (p < 0.001).