Text data mining and customer demand insight for port logistics marketing based on the transformer model

Le résumé fourni par la source

To address the issue of low efficiency and insufficient accuracy of traditional customer demand insight methods in the context of intensified competition in the port logistics industry, this paper focuses on the research of customer demand insight using port logistics marketing text data. Firstly, text data from multiple channels such as enterprise customer management systems and social media over the past 5 years were collected. After preprocessing such as cleaning and tokenization, an improved BERT model's "topic mining + sentiment analysis + demand element extraction" multi-task learning framework was constructed. The model's performance was verified through comparative experiments with traditional LDA models and the basic BERT model. The experimental results show that the F1 values of the improved BERT model in tasks such as customer demand topic mining, sentiment analysis, and demand element extraction reached 89.7%, 92.3%, and 91.5% respectively, which were 18.2%, 21.5%, and 19.8 percentage points higher than those of the traditional LDA model, and 4.3%, 3.1%, and 3.7 percentage points higher than those of the basic BERT model. Based on the mining results, five core demand themes such as efficient transportation and low-cost storage were extracted, and the pain point issue of customers' satisfaction with timely services was identified, which was only 68.3%. The research results provide data support for port logistics enterprises' precise marketing and service optimization, effectively improving the efficiency and accuracy of customer demand insight.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.