WynnDing/StatefulLSTMforecast

在企业场景中，对订单或者发货的预测，通常是一个超级难度的不可控事件；如果依靠利用经典概率算法的深度学习神经网络，那么特征向量猜想及数据收集工作在实际落地过程中将成为不可能；而通过采用仅仅依靠订单或者发货的时间序列数据，通过有状态的LSTM递归神经网络，通过调整时间窗口来训练数据，可以做到数据的回归预测。In the enterprise scenario, the forecast of orders or shipments is usually an uncontrollable event of super difficulty; if relying on deep learning neural networks using classical probability algorithms, the feature vector guessing and data collection work will become Impossible; and by using time-series data that relies solely on orders or shipments, through adopt stateful LSTM recurrent neural network, by adjusting the time window to train the data, regression prediction of the data can be achieved.

Python