Exploring Optimizing Paths for the Allocation Efficiency of Core Production Factors in the AI Economy

Abstract

To address the issues of poor adaptability of fixed weights and premature convergence in the allocation of core production factors in the AI economy, traditional particle swarm optimization (PSO) algorithms are proposed. This paper proposes a dynamic weight adaptive particle swarm optimization (DWAPSO) algorithm. This algorithm constructs a three-layer framework: "state perception - weight adaptation - strategy adjustment." It incorporates scenario parameters such as resource surplus rate and demand urgency into the dynamic weight calculation, designs adaptive perturbations to avoid local optima, and constructs a multi-objective optimization function based on resource utilization, allocation fairness, and economic output. Experiments based on 1,200 sets of AI economic factor data compared with traditional PSO and genetic algorithms (GA) in four typical scenarios showed that DWAPSO achieved an average resource utilization rate of 92.4%, a 14.1 percentage point increase over traditional PSO (78.3%) and a 17.2 percentage point increase over GA (75.2%). The average variance in allocation fairness was 0.08, a 46.7% decrease over traditional PSO (0.15). Economic output growth averaged 23.6%, a 7.8 percentage point increase over traditional PSO (15.8%). DWAPSO also exhibited faster convergence (resource utilization reached 88.2% after 50 iterations) and greater stability (coefficient of variation <0.02), effectively meeting the requirements for dynamic configuration of AI factors.