Do supply chain related factors enhance the prediction accuracy of blockchain adoption? A machine learning approach

This study employs a machine learning approach to examine whether and to what extent supply chain related factors can improve the prediction accuracy of blockchain technology (BT) adoption. The supply chain factors studied include supply chain collaboration, information sharing along the supply chain, partner power, trust in supply chain partners and Guanxi with supply chain partners. We choose the Technology-Organization-Environment (TOE) framework as the benchmark model and quantify the importance of supply chain factors by comparing the prediction accuracy of the benchmark model using only the TOE framework with an extended model combining supply chain factors with the TOE framework. Based on data collected from 629 Chinese firms, we find that Support Vector Machine stands out among all machine learning algorithms: the complete model including both supply chain and TOE factors reaches an accuracy rate of 89.3 %, while the model including only TOE factors has an accuracy rate of 83 %. Based on a 10-fold cross-validated paired t-test, we further confirm that incorporating supply chain factors can significantly improve the prediction accuracy by 6.34 % over the benchmark model. Our results indicate that TOE factors are insufficient to understand and predict BT adoption; supply chain factors also need to be considered.