Assessing Material and Labor Cost Indices Impacts on Construction Firms’ Financial Performance: A Machine Learning Comparison with Feature Engineering

초록

Financial performance serves as the primary metric in project management to make strategic decisions based on internal and external data. While traditional manufacturing sectors utilize these indicators to examine stability, the construction industry possesses distinct characteristics defined by high material reliance and labor-intensive operations. This study employs an interpretable machine learning framework to assess the impact of 11 material and labor indices on the return on assets of 20 major Korean construction firms from 2015 to 2023. The results show that engineered cost features contribute 56.97% to the predictive power, with the interaction between the asset turnover and cement producer price index emerging as the most influential predictor (15.00%). Material-related features (42.01%) substantially exceeded labor-related features (13.07%), reflecting the higher volatility of commodities. Time-lagged analysis confirms a two-quarter delay in the translation of cost changes to financial performance. These findings provide actionable insights into adaptive cost risk management and firm-specific hedging strategies.

키워드