A Used Sailboat Pricing Problem Based on Graph Convolutional Neural Network

Used sailboats change with age and market conditions, and prediction models that extract feature correlations are needed to identify hidden features that affect the listing price of used sailboats and accurately predict the listing price of used sailboats. In this paper, we introduce a graph neural network model and an attention machine mechanism for the problem of predicting the listing price of used sailboats. We collected data on the brand, variety, regional GDP and cargo throughput of sailboats. To abstract information about the dependencies between the two data. On the data side, each sample is used as a node and feature vector (brand, variant, regional GDP, cargo throughput, etc.) to construct the graph data. Treat each sample as a node feature. The similarity between two graph nodes is calculated using Pearson similarity algorithm and used as initialized edge weights. On this model, a graph convolutional neural network (GCN) is constructed with an MSE loss function and the final loss is obtained after bringing the data into the model for training. The attention mechanism is then introduced and the attention weight matrix is output to obtain the weights of the sailboat feature vector, where the average cargo throughput (tons) has the greatest effect on the listed price of the sailboat used.