Prediction of Product Order Demand Based on LSTM Models with Different Time Scales
This article uses LSTM models to forecast the demand for products in four sales regions at daily,weekly,and monthly time scales.The results indicate that when using the daily scale model,the time interval is too long for the model to converge,suggesting the need to shorten the interval when building the model.On the other hand,when using the weekly scale model,the loss function gradually converges to the x-axis,indicating a more significant training effect.The monthly scale model still has room for improvement and requires further training for complete convergence.The choice of different time scales has an impact on the accuracy of the predictions,as very short or very long time scales can increase prediction errors.
NETL
NSTL
万方数据
