Previous studies have focused on explaining the BIM adoption or implementation, and little is known about the diffusion trend. Thus, this study evaluates the BIM diffusion process in SMEs with a predictive machine learning modelling approach using the Nigerian construction industry as a case study. Empirical questionnaire was employed for BIM diffusion data collection, and the collected data was cleaned and balanced with hybridized SMOTE +Tomek links. Ensemble machine learning algorithms were applied to the collected data to develop predictive models for the BIM diffusion process. The optimized and best-performing models were stacked via a new model (ensemble of ensembles) and deployed in an interactive Python-based Application - BIM Diffusion Prediction System (BIM-DPS) - using Streamlit framework. The most important features for BIM awareness in SMEs are observability, top management support, and normative pressure, while the top predictors of intention to adopt BIM in SMEs are compatibility, top management support, and behaviour. These sets of the same features have different impacts on awareness and adoption. Also, these features are from technology, organisation and environment contexts of the TOE framework and underscore the social-technical nature of BIM, which should be reflected in strategies to drive proliferation. The study highlights the strong predictive performance of stacked ensemble models. It provides an easy-to-use application to forecast the behaviour of firms to mitigate risks and develop tailored interventions.
Saka et al. (Tue,) studied this question.