Interbank Offered Rate Forecasting Using PSO-LS-SVM

In the procedure of China´s market-oriented reforms of interest rates, the interest-rate risk becomes increasingly apparent. Analyzing the existing studies and the interbank offered rate trend, this paper finds that LS-SVM, which is short for Least Squares Support Vector Machines, is excel in nonlinear data approximation, which is suitable for interest rate forecasting. Firstly, the paper established a standard LS-SVM model. Secondly, Particle Swarm Optimization is introduced to optimize the parameters of LS-SVM. Thirdly, a standard SVM model optimized by PSO and a BP neural network are established for comparison. Then the experiment is conduct to forecast the offered rate in China´s interbank market. The results show that the PSO-LS-SVM model outperforms any other approaches since its RMSE of testing is 0.04 and the relative errors between forecasting values and the actual ones are all below 0.18, which demonstrates the performance of the PSO-LS-SVM model in interest rate forecasting is promising.