Data-driven modeling of technology acceptance: A machine learning perspective

• Published in: Expert systems with applications Vol. 185; p. 115584
• Main Authors: Alwabel, Asim Suleman A., Zeng, Xiao-Jun
• Format: Journal Article
• Language: English
• Published: New York Elsevier Ltd 15.12.2021; Elsevier BV
• Subjects: Acceptance tests; Algorithms; End users; Evaluation; Information systems; Machine learning; Modelling; Prediction models; Predictions; Predictive analytics; Sensitivity analysis; Structural equation modeling; Support vector regression; Technology acceptance; Technology Acceptance Model; Technology assessment; Technology utilization; UTAUT; Technology acceptance; Support vector regression; Structural equation modeling; UTAUT; Predictive analytics; Machine learning
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2021.115584

•This research highlights the impact of machine learning on technology acceptance.•37 constructs predicted personal technology acceptance within consumer-use context.•Past behavior was found the most influential determinant of technology use.•Support vector regression-polynomial created a unique paradigm improvement.•Sensitivity analysis-partial derivatives improved evaluation of constructs’ ranking. Understanding, explaining, and predicting technology acceptance have dominated the research of information systems (IS) for more than two decades. Past research has favored explanatory modeling, considering it a prediction-oriented approach; until recently, predictive analytics has been poorly understood and widely underappreciated. Research on IS for prediction-oriented modeling like predictive analytics remains rare, despite its potential for development and utility. Our research addresses the capacity of predictive analytics for advancing technology acceptance modeling by assessing predictive power, evaluating the current frameworks, and introducing new constructs. This research formulates a unique data-driven approach that utilizes machine learning (ML) and predictive analytics-based modeling to empirically predict end users’ acceptance of consumer-use technology in a non-organizational setting using the following steps. First, a thorough analysis of IS literature was conducted to explore the constructs of technology use in various contexts. Second, the Twitter API and interviews were utilized to extract new constructs and evaluate the content of current models of technology acceptance. Third, a unique technology acceptance model of thirty-seven constructs was developed and tested on thirty-two personal technologies with heterogeneous subjects. Fourth, ML algorithms estimated the predictive power of the model and ranked the influence of its variables, achieving an R2 of 0.97 and an error rate of 0.04. Thirteen new constructs were successfully introduced, including eight technology characteristics. Four other constructs were reinstated, presenting the utility of the ML approach to contribute to research. Ranking the thirty-seven constructs by applying a sensitivity analysis on the basis of partial derivatives showed the differences between the predictive model and the explanatory model of personal technology acceptance. The proposed approach demonstrates the capacity of ML to formulate a complex model of personal technology acceptance, which further develops technology acceptance models.