Ensemble Sailboat Price Prediction Model Based on GA Optimization

We attempt to establish a corresponding predictive model by collecting data that may be related to the listing price of used sailboats. After preprocessing the data, we selected specific algorithms and algorithm combinations to predict the prices of used sailboats, and then selected specific indicators to evaluate the predictive results of each algorithm. In the experimental process, we first experimented with the Random Forest and XGBoost algorithms for the model. Then we used the genetic algorithm to optimize the hyperparameters of the Random Forest and XGBoost algorithms, obtaining the GA-RF and GA-XGBoost algorithms, which showed improved performance. In addition, we used the stacking method to stack the RF and XGBoost algorithms to obtain a mixed model. The stacked model had high training efficiency and its prediction accuracy was comparable to that of the GA-tuned algorithms. To further improve accuracy, we optimized the stacked model and named it the GA-RF-XGBoost algorithm. This algorithm applied the three machine learning methods of Bagging, Boosting, and Stacking, as well as the GA optimization algorithm. To compare results, we also combined the Random Forest and XGBoost algorithms, which were not optimized by the genetic algorithm, using the stacking method to obtain the RF-XGBoost algorithm. The final experimental results showed that the integrated optimization model obtained by stacking RF and XGBoost and then optimizing with the GA algorithm performed the best in predicting the listing price of used sailboats, but it also required the most training time and was less efficient than simply stacking RF and XGBoost.