A web text mining approach for the evaluation of regional characteristics at the town level

Abstract

AbstractThe evaluation of regional characteristics can reveal the advantages of local industries that are significant to guide regional industrial restructuring and industrial layout. Detailed guidance requires fine‐grained evaluations of the regional characteristics, especially at the town level, since towns are the locations of industry. Existing evaluation methods, such as the location quotient and the Porter diamond model, depend on statistical data to compare the advantages of regional characteristics in different regions. Statistical data have fixed statistical items and spatial units that limit the content and granularity, respectively, of the evaluation. In existing methods, non‐covered industries and under‐counted units, especially at the town level and below, result in incomplete descriptions of the regional characteristics. In contrast, web text in the current internet era contains numerous descriptions of regional characteristics. Therefore, web text can potentially be used to evaluate these characteristics. This article proposes a novel web‐based method for the evaluation of regional characteristics (WERC). According to the features of the regional characteristics of the town in the web texts, the WERC method uses the term frequency method to extract the typical characteristics of the region by crawling text on websites and compares the relative advantage of the typical characteristics between different regions to determine the outstanding regional characteristics. WERC is used in a case study to evaluate the regional characteristics of 1,090 towns in Fujian Province, China, obtaining high precision (0.946) and recall (0.895). Unlike existing methods, the proposed method provides comprehensive quantitative town portraits to describe the regional characteristics. The town portrait not only shows the advantages of the typical characteristics of the town, but also quantifies the advantages using a ranking of the typical characteristics. The results can guide regional industrial restructuring and industrial layout and provide a novel approach for the evaluation of regional characteristics based on web text.