A two-phase TOE framework integrating SEM and ANN for evaluating cloud computing adoption in software testing

Context:
Software applications are increasingly complex, distributed, and component-based, making software testing more demanding-especially as large-scale projects generate thousands of test cases. Cloud computing adoption (CCA) has become essential in such scenarios to ensure scalability, efficiency, and resource optimization.

Problem Statement:
However, existing research on cloud-based testing (CBT) often relies on linear methods, failing to capture nonlinear adoption dynamics. Hybrid models like structural equation modeling (SEM)-artificial neural networks (ANNs) and empirically validated Technology–Organization–Environment (TOE) frameworks are underutilized. Cross-industry variations, evolving adoption trends, and key factors like thresholds and moderating variables remain largely unexplored, creating significant gaps in understanding CBT adoption behavior.

Objective:
The aim of the study is to assess CCA for CBT by introducing a novel two-phase TOE framework combining SEM and ANN.

Methods:
A Systematic Literature Review (SLR) of 136 papers was conducted to identify seventy motivating factors (MFs) related to CBT. These MFs were then categorized into ten predictors, and based on these predictors ten hypotheses were developed. A two-phase model based on SEM and ANN was developed to test the hypothesis.

Results:
Out of ten hypotheses, eight were supported. Five predictors i.e. perceived degree of cloud resource utilization, perceived economic benefits of CC adoption, perceived level of trust in cloud adoption, perceived external stimuli and feasibility planning and risk analysis positively influenced CBT adoption. Three predictors: perceived business concerns, perceived effort expectancy, and organizational competence and capacity had a negative effects. Two predictors, perceived performance expectancy and organizational dynamics and business drives, showed no significant impact.

Novelty:
Some studies have been conducted to examine the MFs affecting CCA but they are not specifically conducted for software testing. Moreover, no attempt was made to explore their multifaceted linear and non linear affect on CCA.

Conclusion:
This study suggests that software testing should be conducted in the cloud environment.