Used sailboat price prediction based on GBRT algorithm

Brokers often face the situation of diverse variations and too many factors to consider when pricing used sailboats, which may lead to subjective influences. Prediction of complex problems is difficult because the direct relationship between used sailboat prices and their associated characteristics is often not linear, and the relationship between prices and characteristics is difficult to capture. Also, the price movement patterns of monohull and catamarans sailboats are often different and need to be considered separately when forecasting the prices of both types of sailboats. To this end, we combined market data and machine learning to develop a GBRT(Gradient Boosting Regression Tree)-based used sailboat price evaluation model, aiming to assist brokers in pricing the used. The experiments show that the prediction model built based on the GBRT algorithm has better prediction accuracy. Further, MSE, MAPE, and R
²
were used to quantify the prediction performance, and the results showed that the GBRT algorithm also has better prediction performance compared with the multiple linear regression algorithm, BP neural network algorithm, and XGBoost algorithm.