Personalization, the de facto standard for modern recommendation systems, requires user profiling based on item interactions. However, existing models struggle to condense extensive user preferences into a single representation, leading to computational inefficiencies. To address these, this paper proposes the Symbolic-Regression Based User Profiling (SRBUP) framework, which represents user profiles using just two parameters: slope and intercept, forming symbolic objects that reduce storage and computational complexity. Similarity between users is then measured using the area (A) between those symbolic regression lines and behavioral (B) differences they convey. These similarity scores are leveraged to generate recommendations. Beyond profiling, generating relevant recommendations is another key focus of this paper. Towards this, four recommendation strategies with hierarchical intersections are devised, each examining the impact of different thresholds on predicted ratings. Experimental results show that selecting an optimal threshold depends on balancing quality, diversity, and consensus filtering. A comparative analysis using Normalized Discounted Cumulative Gain (NDCG) and Mean Reciprocal Rank (MRR) confirms that SRBUP outperforms existing models in ranking quality and relevance.