Comparison of Cost Prediction Machine Learning Model for Early Stages Warehouse Steel Structure Project using XGBoost dan Random Forest

Purpose: The objective of this paper is to review machine learning prediction models to predict warehouse steel structure project cost in early stages with limited data using extreme gradient boosting (XGBoost) and random forest (RF). This cost prediction model aims to minimize predicted cost error from actual cost thus improve accuracy estimation cost in early stages. Design/methodology/approach: This paper identifies cost driver variables that affects estimated cost of warehouse project with a total of 15 variables identified. Case study consisting 32 warehouse steel structure project cost data collected from construction company in Indonesia were used to train and test cost prediction model using Extreme Gradient Boosting (XGBoost) and Random Forest (RF). The accuracy of the predicted cost was calculated using mean absolute percentage error (MAPE), mean squared error (MSE), R-squared (R2) and root mean squared error (RMSE). The results from both models were compared to find which models produced most accurate cost prediction. Findings: This study reveals that with limited data RF model produced more accurate cost predictions with MAPE 3.50% compared to XGBoost model with MAPE 4.03%. Models also discover total area of floor, office area and steel material cost as the most affected variables to the predicted costs. Research limitations/implications: This study’s shortcoming are the samples used to case study were small and limited to warehouse steel structure type buildings. The variables also limited to that can be obtained in early project phase. Originality/value: This study contributes in reviewing results of machine learning prediction models for warehouse steel structure project cases. It provides alternative cost estimation methods and also help identifies which variables most affected cost that helps in decision making process for which variable to focus on making more efficient warehouse steel structure project cost.