Construction Project Cost Prediction Based on Genetic Algorithm and Least Squares Support Vector Machine

For the small sample data and complex nonlinear characteristic of construction project cost, a new hybrid prediction model combing genetic algorithm and small sample learning model based on least squares support vector machines is proposed. First, all the candidate features are ranked by correlation with the dependent variable, the front ranking features are used to initialize part of population for the genetic algorithm to get better feature subset, and then the construction cost prediction model of the least square support vector machine is constructed. Experiments on Jiangsu Province housing project data show an improved performance over other models in prediction accuracy, it is an effective method of project cost forecasting.