Research On Merger Waves Prediction And Link Prediction Algorithms Based On Supply Network

With the intensification of globalization and market competition,corporate merger and acquisition activities are continuously common worldwide.Meanwhile,predicting supply relationships has become increasingly important in supply management.In this context,researchers and business professionals have been exploring how to use supply network information to predict merger waves and possible links,and to help companies better understand the risks and opportunities in the market.In recent years,with the rise of social networks in the field of data science,various derivative networks such as author collaboration networks,citation networks,and user behavior networks have been discovered and applied.Supply networks also have their important significance in studying corporate behavior and have been applied in risk management,bankruptcy prediction,and other areas.However,the research on enterprise supply networks is relatively scarce,and there is still potential to be explored.Regarding the current situation in the field of enterprise relationship networks,this thesis focuses on using the supply network to conduct predictive activities within the industry:1.A method for predicting the industry’s merger and acquisition wave is proposed in this thesis.This method relies on a dataset containing financial statements and supply information,as well as merger and acquisition records of over 60,000 companies.Based on these data,a 1000-dimensional dataset is constructed,including financial descriptive data and network structure data.Different machine learning algorithms are used to build classifiers for prediction.The experimental results show that compared with traditional merger waves prediction methods,this method has a significant improvement in precision-the prediction precision of the acquirer reaches 91%,and the prediction precision of the target reaches 96%.At the same time,the addition of supply network structural features also improves predictive performance,especially by analyzing the micro-structure network characteristics,new factors that affect the merger and acquisition wave are discovered.Further analysis shows that the features that play a decisive role in prediction fit with the well-known financial theory,Tobin’s Q theorem.2.In the constructed supply network for automotive enterprises,a path-based node representation method is proposed to predict the link between nodes from a similarity perspective.This method combines multiple relationships and nodes,and differs from traditional similarity measurement methods,providing better performance for link prediction on heterogeneous bipartite networks.In scenarios where the data is not sufficient,it can maximize the role of the supply network.Further analysis of the automotive supply network in this thesis revealed the existence of communities,and the performance of link prediction within the community was further improved,which has important reference significance for other supply networks.