Study of Food Cold Chain Logistics Demand Forecast Based on Multiple Regression and AW-BP Forecasting Method on System Order Parameters

The food cold chain logistic demanding is increasing in China, while the theoretical research is relatively slow. This gap between theory and practice leads to great social conflicts. The government is in urgent needs of mastering the future development scale of food cold chain logistic system to achieve scientific planning and guiding and to prevent a series of consequences of blind construction. As the forecasting work has difficulties with the lack of relevant theories and statistical data, conventional forecasting methods are not applicable here. The Multiple Regression and AW-BP forecasting method based on system order parameters is designed and applied, considering that the food cold chain logistics is a complex nonlinear system, and as well as the limitation and availability of the statistical data. The forecasting methods are based on the Dynamic System Theory, establishing the system order parameters of cold chain logistics scientifically, making full use of the comprehensiveness of the multiple regression and the nonlinearity of the BP neural network, minimizing the negative impacts of lack of data on prediction, meanwhile correcting the deficiencies of general BP neural network as slow convergences and being easily trapped in local optimum by introducing the correction of error function and the dynamic adaptive weight. It is proved through many experiments that the convergence speed, prediction accuracy and local extremum avoidance of new forecasting method have been greatly improved. For cold chain logistics, the new forecasting method is adaptable, simple, useful and efficient, and is worthy of being generalized.