China has experienced rapid industrial land growth over last three decades, which has brought about diverse social and environmental issues. Hence, it is extremely significant to monitor industrial land and intra-structure dynamics for industrial land management and industry transformation, but it is still a challenging task to effectively distinguish the internal structure of industrial land at a fine scale. In this study, we proposed a new framework for sensing the industrial land and intra-structure across the urban agglomeration around Hangzhou Bay (UAHB) during 2010–2015 through data on points of interest (POIs) and Google Earth (GE) images. The industrial intra-structure was identified via an analysis of industrial POI text information by employing natural language processing and four different machine learning algorithms, and the industrial parcels were photo-interpreted based on Google Earth. Moreover, the spatial pattern of the industrial land and intra-structure was characterized using kernel density estimation. The classification results showed that among the four models, the support vector machine (SVM) achieved the best predictive ability with an overall accuracy of 84.5%. It was found that the UAHB contains a huge amount of industrial land: the total area of industrial land rose from 112,766.9 ha in 2010 to 132,124.2 ha in 2015. Scores of industrial clusters have occurred in the urban-rural fringes and the coastal zone. The intra-structure was mostly traditional labor-intensive industry, and each city had formed own industrial characteristics. New industries such as the electronic information industry are highly encouraged to build in the core city of Hangzhou and the subcore city of Ningbo. Furthermore, the industrial renewal projects were also found particularly in the core area of each city in the UAHB. The integration of POIs and GE images enabled us to map industrial land use at high spatial resolution on a large scale. Our findings can provide a detailed industrial spatial layout and enable us to better understand the process of urban industrial dynamics, thus highlighting the implications for sustainable industrial land management and policy making at the urban-agglomeration level.

PurposeThe focus of industrial cluster innovation lies in the cooperation between enterprises and universities/scientific research institutes to make a theoretical breakthrough in the system and mechanism of industrial cluster network. Under the theoretical framework of cluster network, industrial structure can be optimized and upgraded, and enterprise benefit can be improved. Facing the increasing proliferation and multi-structured enterprise data, how to obtain potential and high-quality innovation features will determine the ability of industrial cluster network innovation, as well as the paper aims to discuss these issues.Design/methodology/approachBased on complex network theory and machine learning method, this paper constructs the structure of “three-layer coupling network” (TLCN), predicts the innovation features of industrial clusters and focuses on the theoretical basis of industrial cluster network innovation model. This paper comprehensively uses intelligent information processing technologies such as network parameters and neural network to predict and analyze the industrial cluster data.FindingsFrom the analysis of the experimental results, the authors obtain five innovative features (policy strength, cooperation, research and development investment, centrality and geographical position) that help to improve the ability of industrial clusters, and give corresponding optimization strategy suggestions according to the result analysis.Originality/valueBuilding a TLCN structure of industrial clusters. Exploring the innovation features of industrial clusters. Establishing the analysis paradigm of machine learning method to predict the innovation features of industrial clusters.