基于FOA优化混合核LSSVM的铁路货运量预测

单一核最小二乘支持向量机(LSSVM)在铁路货运量预测中难以准确描述货运量的复杂变化特征，限制了预测精度的提高。针对该问题，提出一种基于果蝇算法(FOA)优化混合核LSSVM的预测方法。以多项式核与径向基核组合的混合核函数作为LSSVM核函数，构建铁路货运量的混合核LSSVM预测模型，同时利用FOA全局寻优能力强、计算速度快等优点优化选择混合核LSSVM参数。以我国铁路货运量为例进行方法验证。结果表明，所提方法的RMSE、MAE、MAPE和THEIL值分别为8433.0、6670.8、0.0180和0.0117，均小于其他模型，FOA算法搜索混合核LSSVM参数的时间为40.2948秒，分别比GA和PSO算法减少了2.6208秒和20.7016秒，适合于铁路货运量的短期预测。

Forecast on Railway Traffic Volume Using Mixed-Kernel LSSVM Optimized by FOA

It is difficult for the single-kernel Least squares support vector machines (LSSVM) to describe accurately the complexity change feature of volumes in railway freight volume forecasting, which limits the improvement of forecasting accuracy. To solve the problem, this paper proposes a new forecasting method based on fruit fly optimization algorithm (FOA) and mixed-kernel LSSVM. First, the mixed-kernel LSSVM is constructed for railway traffic volume forecasting, in which the mixed kernel function that linearly combines polynomial kernel and radial basis kernel is used as the kernel function of LSSVM. Second, FOA is employed to optimize the parameters of the mixed-kernel LSSVM based on the advantages of the global searching ability and quick computing speed. Finally, the railway traffic volume of China is used to prove the effectiveness of the provided method. The results show that the value of RMSE, MAE, MPE and THEIL of the proposed method are 8433.0, 6670.8, 0.0180 and 0.0117, respectively, being less than the other methods. The time for searching the optimal parameters in the mixed-kernel LSSVM by FOA is 40.2948s, reducing 2.6208s and 20.7016s relative to genetic algorithm (GA) and particle swarm optimization (PSO) algorithm. The proposed method is applicable to forecasting short-term railway freight volumes.