A neural network approach to predicting the net costs associated with BIM adoption

Abstract

A neural network approach is proposed to estimate the costs and benefits associated with implementing Building Information Modelling (BIM) at firms. This includes specifying the BIM applications and the resources required for reaching a specific level of detail within the generated models, referred to as level of development (LOD). Such predictions are imperative to decision makers and can aid in the examination of the best strategies to execute when deciding on the adoption and implementation of BIM. The proposed neural network is customised to suit a firm's investment plan when it comes to BIM implementation. Multi-label and multi-class classifications are adopted to derive the cost and benefit functions for BIM application and LOD implementation, respectively. Threshold functions to distinguish the positive and negative labels in multi-label classification are adopted. The proposed neural network is developed based on data collected from Australian and Chinese construction firms using a 7-point Likert type questionnaire. The proposed neural network provides decision-makers with a tool to assess which BIM/Non-BIM applications to implement, along with the LOD that is most suited to the organisation's financial and technical ability.