iCHAC: A Deterministic Tensor-Based Constraint-Aware Framework for Combinatorial Resource Allocation

Abstract

Combinatorial resource allocation problems are NP-hard and arise in various domains such as academic timetabling and cloud resource management. Traditional exact optimization methods ensure optimality but suffer from poor scalability, while metaheuristic approaches often experience stochastic instability, high computational overhead, and reproducibility challenges. This paper proposes iCHAC, a deterministic tensor-based framework for solving multi-constraint allocation problems. The proposed method models resource conflicts as vectors within a multi-dimensional resemblance tensor, enabling simultaneous constraint evaluation through vectorized linear algebra operations. By replacing iterative constraint verification with vectorized agglomeration, the algorithm reduces computational complexity while guaranteeing feasibility by construction. An expanded deterministic strategy bank is incorporated to mitigate local optima without relying on randomness. Experimental evaluation of timetabling dataset from the College of Industrial Technology in Misrata (CIT) demonstrates that iCHAC achieves high feasibility, improved computational efficiency, and fully reproducible results compared with traditional optimization and metaheuristic approaches.