A study published in Scientific Reports has built an integrated framework combining a BERT neural network with regional economic analysis to quantify how ice-snow consumption links different regions across China, using Harbin Ice-Snow World as its central case. The researchers, Yanqi Liu and Shibo Zhi, integrated semantic features extracted from approximately 12,000 consumer reviews with multidimensional statistical data drawn from 15 regional nodes, constructing a regional synergy intensity model designed to measure the strength of relationships between places tied together by winter tourism spending.

The team validated the approach in two ways before reporting results. Five-fold cross-validation assessed the model’s predictive reliability, while a three-expert consistency assessment of the underlying annotations produced a Fleiss’ Kappa of 0.84, indicating substantial agreement among the human raters. In predicting synergetic relationships, the model achieved an F1-score of 0.87, which the authors report significantly outperformed baseline models built on advanced large language models such as Llama-3-70B as well as ablation variants of their own framework.

Network analysis of the results identified Harbin as the “absolute core” of the regional system, with a weighted centrality score of 4.52. Using quantitative topological indicators, including network density and path strength, the researchers classified the synergetic patterns into three distinct zones: a “Heilongjiang Core Agglomeration Zone” centered on the host province, a “Jilin-Liaoning-Hebei Functional Complementarity Zone” in which neighboring provinces appear to play mutually supporting roles, and a “Long-Distance Characteristic Linkage Zone” connecting more distant regions. These categories describe the observed spatial organization of the ice-snow industry as captured by the model, rather than predicted future outcomes.

The authors position the work as offering a new perspective for understanding the spatial organization of the ice-snow industry and as providing data-driven decision support for formulating cross-regional collaboration policies. As a case study centered on a single flagship attraction and its associated regional nodes, the framework’s findings are specific to the ice-snow consumption network it was trained on, and the synergy zones represent model-based classifications of observed patterns. The study was published on 22 October 2026 as an open-access article, shared early as a citable accepted version carrying a permanent DOI, with the final Version of Record to follow after further edits.

Subject of Research: Regional synergy analysis of ice-snow consumption using a BERT neural network

Article Title: Regional synergetic effects of ice and snow consumption based on a BERT neural network

Article References: Liu, Y., & Zhi, S. (2026). Regional synergetic effects of ice and snow consumption based on a BERT neural network. Scientific Reports. https://doi.org/10.1038/s41598-026-77353-4

Image Credits: AI Generated

DOI: 10.1038/s41598-026-77353-4

Keywords: BERT, neural network, ice-snow tourism, Harbin Ice-Snow World, regional synergy, consumer reviews, network analysis, Scientific Reports, large language models, cross-regional policy, Regional, synergetic

News Source: Cassandra Pierce. (October 24, 2026). BERT neural network maps regional synergy in China’s ice-snow economy. Scienmag.