Deep Learning Model for Life Cycle Cost Assessment in Industrial Buildings: A DBN-Based Approach for Structural Decision-Making

To manage industrial building construction and operation efficiently, it is necessary to perform the life cycle cost assessment, which will help evaluate the long-term financial consequences. Previous techniques used in Life Cycle Cost Assessment (LCCA) are useful but are constrained by reliability, speed, and management of large data sets. This paper presents an improved methodology for increasing the exactness of the Life Cycle Cost Assessment Life Cycle Cost Assessment (LCCA) for industrial buildings using the Deep Belief Network (DBN). The Deep Belief Network (DBN) model, using Python software, uses the multiple layers of the Restricted Boltzmann Machines and makes an increasing level of hierarchy capture of cost data patterns and correlations. This approach eliminates the limitations characteristic of the more traditional approaches by enabling the user to provide and/or work with more accurate predictive models of the nonlinear relations between the data sets and the targets. The model was trained using a two-phase process involving unsupervised pre-training and supervised fine-tuning, leading to improved performance metrics compared to traditional regression-based models. The Deep Belief Network (DBN) model achieved a Mean Absolute Error (MAE) of 2.8%, a Root Mean Squared Error (RMSE) of 3.2%, and an R-squared (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) value of 0.96, indicating a high level of accuracy and fit. Explained in detail are various cost parameters and assessment of life cycle cost estimates used in the model, results that depict the model as highly accurate in estimating life cycle costs with the potential to greatly improve on decision-making processes of industrial building management. Subsequent investigations will relativize the biological model with other data sources and/or combine them to further enhance the model and the range of its applications.