A complementary SEM and deep ANN approach to predict the adoption of cryptocurrencies from the perspective of cybersecurity

Abstract

The present study investigated the role of six fundamental attributes of cybersecurity (i.e., authenticity, availability, confidentiality, integrity, possession/control, and utility) in predicting the adoption of cryptocurrencies. The study developed a prediction model and evaluated this model using a complementary approach by integrating structural equation modeling (SEM) and a deep artificial neural network (ANN) model. The sample of the study consisted of 450 cryptocurrency users aged between 18 and 38. The SEM results showed that availability, integrity, utility, and possession/control significantly predict attitudes, which in turn significantly predict continuous intention to use cryptocurrencies. The paths specified in the structural model accounted for 24% and 85% of the variance in attitude and continuous intention, respectively. Furthermore, the prediction model was tested by using a deep ANN multi-layer perceptron (MLP) algorithm. The algorithm predicted the attitude with a mean accuracy of 60.59% and 66.82% for training and testing, respectively. The result indicated that the deep ANN performed better than SEM in predicting attitude. The complementary approach enabled the discovery of both nonlinear and linear relationships between the variables and thereby contributed to accurately predicting adoption behavior.