AN EARLY WARNING MODEL OF ROAD ENGINEERING BIDDING RISK BASED ON IMPROVED RANDOM FOREST

Lots of challenges, such as low precision and low efficiency in road project bidding risk warnings, have arisen in the previous years. Moreover, such problems have caused several mortal accidents. Therefore, this paper proposes a risk warning model for road project bidding based on an improved random forest algorithm. The methodology consists of several layers: first, this paper identifies the early warning factors that affect road engineering bidding. Then, an early risk warning model of road engineering bidding is constructed based on random forest. As the bidding data of road engineering is typically unbalanced, a Synthetic Minority Oversampling Technique (SMOTE) is used to expand the data of the sample training set to reduce its deviation. Finally, this paper selects the reconstruction and expansion project of Lidong, Jiangxi Province, China as a case study where it shows that, compared with traditional machine learning algorithms (artificial neural network, support vector machine, etc.), the proposed model has the advantages of strong generalization performance, simplified par rameters, and modeled structure.