Expediting the Cost Estimation Process for Aged-Housing Renovation Projects Using a Probabilistic Deep Learning Approach

Abstract

Since the early 1980s, the Korean government has rapidly boosted residential buildings to cope with substantial housing shortages. However, as buildings have been aging simultaneously, the performance of a large number of residential buildings has deteriorated. A government plan to upgrade poor housing performance through renovation is being adopted. However, the difficulty of accurate construction cost prediction in the early stages has a negative effect on the renovation process. Specifically, the relationship between renovation design elements and construction work items has not been clearly revealed. Thus, construction experts use premature intuition to predict renovation costs, giving rise to a large difference between planned and actual costs. In this study, a new approach links the renovation design elements with construction work items. Specifically, it effectively quantifies design factors and applies data-driven estimation using the simulation-based deep learning (DL) approach. This research contributes the following. First, it improves the reliability of cost prediction for a data-scarce renovation project. Moreover, applying this novel approach greatly reduces the time and effort required for cost estimation. Second, several design alternatives were effectively examined in an earlier stage of construction, leading to prompt decision-making for homeowners. Third, rapid decision-making can provide a more sustainable living environment for residents. With this novel approach, stakeholders can avoid a prolonged economic evaluation by selecting a better design alternative, and thus can maintain their property holdings in a smarter way.