Highway freight cost forecast model based on GRA-IPSO-SVR

ObjectiveThe study aims to forecast the highway freight cost accurately, improve the highway freight order transaction rate, solve the problems of fuzzy pricing standard and information asymmetry between shipper and actual carrier. The highway freight coat forecasting method, using grey relation analysis (GRA) combined with improved particle swarm optimization (IPSO) and support vector regression (SVR) for highway freight order, was proposed.MethodFirst, using the Shunfeng Express freight order dataset, the K-nearest neighbour algorithm was used to fill in the missing values of data. The classification features in freight order dataset were converted into numerical features, and the data were normalized to eliminate the influence of different magnitudes among features, so as to complete the pre-processing of data. Next, GRA was used to calculate the grey relation of different features on highway freight prices. The features with grey relation greater than 0.8 were selected as the main influencing factors. Then, considering the problem that IPSO tended to fall into local optimum too early, the non-linear decreasing inertia weights and asymmetric learning factors were introduced to balance the local search and global search ability of IPSO, achieving achieve an improvement on the standard particle swarm optimization. Finally, IPSO was used with the decision coefficient as the objective function. The penalty factor, kernel function coefficient, and insensitive loss function in SVR were iteratively searched. The main influencing factors of highway freight cost were taken as the inputs of IPSO-SVR to finally obtain the forecast result of highway freight cost.ResultIPSO-SVR has the lowest mean absolute error, root mean square error, mean absolute percentage error and the highest coefficient of determination, i.e., 117.87, 332.20, 0.12 and 0.98 respectively, compared with PSO-SVR, SVR and BPNN.ConclusionThe proposed model has good predictive ability and can provide effective method support for highway freight pricing.