A Local LLM -Based Recommender System for Enhancing Air Cargo Performance Through Digital Marketing Strategies

Faced with increasing operational complexity and heightened service expectations, the air cargo industry must adopt intelligent, adaptive decision-support technologies. This study introduces an advanced offline recommender system grounded in a locally deployed large language model (LLM), optimized through Quantized Low-Rank Adaptation (QLoRA). Designed to enhance key logistics and marketing performance indicators, the system demonstrates considerable gains in load factor prediction accuracy and customer acquisition efficiency. Comparative analysis with conventional recommender systems —namely collaborative and content-based filtering —reveals a ۲۳% improvement in predictive accuracy and a ۱۷% reduction in marketing expenditure per booking. This paper articulates the model architecture, performance metrics, and its potential to transform data-driven air cargo operations.

Keywords:

Air Cargo , Recommender System , Local LLM , QLoRA , Digital Marketing