基于遗传算法优化SVR的区域物流需求预测

Forecasting of Regional Logistics Demand Based on Genetic Algorithm Optimized SVR

指出城市物流需求的影响因素多,具有非线性、随机性和样本量有限等特点,因此选择SVR算法处理这类回归问题时具有优势。基于SVR的性能依赖于关键参数选择的特征,利用遗传算法对支持向量回归模型的惩罚参数、核函数参数和不敏感损失函数进行寻优,使用优化后的参数建立支持向量回归预测模型。以武汉市为例,对武汉市货运数据进行实证研究,预测结果验证了模型的可行性和有效性。

In this paper, considering that the performance of the SVR depends on the characteristics of key parameter selection, we used the genetic algorithm to optimize the penalty parameter, kernel function parameter and insensitivity loss function of the support vector regression model, after which, the support vector regression model was supported in the parameterization of the optimized SVR. Next, in the case of Wuhan, we had an empirical research on the freight transportation statistics of the City, which demonstrated the feasibility and validity of the model.