Random Forest for Price Prediction in Civil Construction Stages

Abstract

The main aim of this research is to present the development and validation of a supervised learning model, using Random Forest algorithms, aiming the unit price prediction in civil construction projects and their stages. This approach emerges from the need to improve the accuracy of cost estimates in civil engineering works, considering that the process of procuring materiel prices to develop a quotation faces limitations due to market instability and the complexity of historical data. It was hypothesized that applying supervised learning algorithms could generate more accurate and efficient predictions compared to traditional price acquisition processes. This hypothesis was tested through an experimental design with comparative groups: one that applied the traditional price acquisition process and another that used the developed predictive model. The results showed a reduction in the error percentage from 11.61% to 6.18%, a similar decrease in the time required for quotation development from 522 to 199 seconds, and an increase in accuracy from 89.6% to 94.2%, with statistically differences (p<0.001). Consequently, the conclusion is reached that the use of prediction-based models represents an effective and adaptable tool for cost management in civil works in the province of Trujillo, Peru.