Abstract:The construction project decision-making stage is lack of information, accurate and efficient cost prediction is the key to scientific decision-making. In order to improve the accuracy of pre-project construction cost prediction, it is very important to discusses how to use the big data of historical projects and machine learning to predict the cost of new construction projects. Firstly, the main influencing factors of the cost in the decision-making stage of construction engineering were determined through literature research. Then, the artificial bee colony algorithm (ABC) was used to optimize the support vector machine (SVM) parameters, namely penalty factorand kernel function parameter. Finally, the construction cost prediction model based on ABC-SVM was built. And then, 84 construction projects from a construction cost data platform were used as data sources for model validation. The results showed that, compared with GRID-SVM model and BP neural network model, the ABC-SVM model has higher prediction accuracy and better applicability.