TECHNOLOGY ACCEPTANCE: A MACHINE LEARNING PERSPECTIVE

Abstract

Machine learning (ML) modelling techniques are promising approaches in technology acceptance modelling research. Among other capabilities, ML can be used to advance and validate theories and evaluate their underlying methodologies and predictive power. This research applies linear and non-linear ML supervised regression techniques to formulate a predictive personal technology acceptance model (PTAM) that avoids the drawbacks and limitations of current technology acceptance models, namely the technology acceptance model (TAM) and the unified theory of acceptance and use of technology (UTAUT). We followed the CRoss-Industry Standard Process for Data Mining (CRISP-DM) to build a novel technology acceptance model that is more generic and predictive than either TAM or UTAUT. We collected data from the technology acceptance literature, tweets (using the Twitter API), interviews and surveys. We used five linear and non-linear regression algorithms to formulate PTAM, which includes thirty-seven features, and applied partial-derivatives sensitivity analysis to rank these features. We also applied Mamdani fuzzy inference to create a fuzzy inference PTAM (FIPTAM) capable of defuzzifying PTAM’s output (use behaviour) values to improve the model’s practicality. We used Bayesian networks to formulate a structured PTAM and create a personal technology acceptance index (PTAI). We then applied data mining techniques to discover new relationship patterns between PTAM’s features and target use behaviour. The resulting PTAM, following CRISP-DM, had better predictive power (R² = 0.97) than either TAM or UTAUT (R² = 0.67 and 0.73, respectively). Past behaviour was found to be the best predictor in PTAM. FIPTAM showed acceptable predictive power (R² = 0.41) and Bayesian networks achieved acceptable accuracy (49.13%). Three types of relationship patterns (non-linear, monotonic and non-monotonic) were discovered among PTAM’s features and target use behaviour. This research demonstrates the capacity of ML techniques to advance technology acceptance frameworks by enhancing their performance, expanding their models and evaluating the relevance of their features.

Date of Award	1 Aug 2022
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Xiaojun Zeng (Supervisor) & Ke Chen (Supervisor)