Firstly, history sales data were fitted by Newton interpolation and neural network was trained by mass data, which makes up for the deficiency of history data.

• 首先采用插值法拟合历史销售数据并求出大量的数据训练神经网络，弥补了历史数据缺乏的问题；