An evolutionary prediction model for enterprise basic research based on knowledge graph

Area of Science:

• Scientific research management
• Innovation studies
• Data science

Background:

• China's enterprise basic research lacks systematic guidance and has dispersed themes.
• Constructing an enterprise basic research knowledge graph is crucial for tracking frontier technology and driving innovation.

Purpose of the Study:

• To develop a predictive model for enterprise basic research evolution.
• To construct an enterprise basic research knowledge graph for better guidance and innovation tracking.
• To identify future hotspots in enterprise basic research.

Main Methods:

• Construction of an enterprise basic research dataset.
• Development of a multilayer Convolutional Neural Network- (CNN-) Bidirectional Long Short-Term Memory- (BiLSTM-) based evolutionary prediction model.
• Inference and complementation of the enterprise basic research knowledge graph.
• Creation of a probabilistic computational model with a multi-attention mechanism to predict future research hotspots.

Main Results:

• The proposed KG-CNN-BiLSTM model significantly outperforms existing classical models in prediction accuracy, AUC, and F1 value.
• The model accurately captures various types of cutting-edge research topics.
• Future hotspots in enterprise basic research were computationally obtained.

Conclusions:

• The developed model provides effective algorithmic guidance for predicting the development trends in enterprise basic research.
• This approach enhances the systematic guidance and focus of enterprise innovation efforts.
• The study offers a robust method for identifying and tracking emerging research frontiers.