The Study and Application of Combination Forecasting Model for Third Party Logistics Demand Based on Multi-Factor Fusion

Under the explosive growth of e-commerce orders, making the accurate prediction of third-party logistics demand is a key research direction. Existing traditional demand forecasting models are generally based on time series models, which only considers a single impact factor and there is less prediction of third-party logistics orders. Therefore, this paper presents a method for combination forecasting third party logistics demand based on multi-factor fusion. First, a multi-factor feature system is built integrating Baidu search index and e-commerce order information. We use MLP multi-layer neural network, AdaBoost decision tree, and support vector machine (SVM) as the prediction models and fuse them into a combined forecasting model by Shapley method. The proposed method is applied to predict the third-party warehousing demand of an e-commerce platform between 2017 to 2018. Results show that the prediction error of this method is below 10% and considerably improves precision.