An Intelligent Prediction for the Innovation Capability of BIM Based on MSCNN-AMIGRU in Industrial Cyber Physical Systems

Building information modeling is a new industrial technology, which is gradually becoming an important part of economic development. How to predict industrial innovation capability has become an important topic. Industrial cyberphysical system (ICPS) is a multidimensional intelligent system cyber communication, calculation, and control. This article combines it with deep learning to propose an industrial innovation capability prediction model via multiscale convolutional neural network and improved gate recurrent unit (IGRU) with attention. First, this uses industrial cyberphysical system to collect relevant industrial data. Second, to extract the complementary spatial features of different scales, this article uses different convolution kernels to build a multiscale fusion network. Third, to fully exploit dependence from each series of data on the time dimension, this article builds an IGRU and combines this module with the attention mechanism (AM). Specifically, the improved model adds 1-tan
h
function operation to reset gate. This can effectively retain the dependence between long-term and short-term information. To obtain more comprehensive feature information, this utilizes an AM to enhance the feature. Finally, the proposed method is tested in many aspects, the results verify the feasibility. The method achieves 95.1%/97.7% accuracy, 92.2%/94.1% recall, and 93.7%/96.2% F1 score.