Commodity order forecasting method and device, storage medium, and terminal

Abstract

A commodity order forecasting method and device, a storage medium, and a terminal are provided, that method comprise the following steps: determining a current width neural network characteristic anda current depth neural network characteristic; the current width neural network feature and the current depth neural network feature are inputted into a width-depth combination model to obtain a predicted order number of a commodity. The scheme of the invention can fully utilize the characteristic information obtained from the historical order big data to improve the accuracy of the commodity order prediction.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to a commodity order forecasting method and device, a storage medium, and a terminal. Background technique [0002] In modern large-scale logistics systems, accurate forecasting of commodity orders has strong economic and social significance. Accurate commodity order forecasting can not only prevent risks such as liquidation and order delays, but also reduce waste of resources in daily operations and improve operational efficiency. [0003] As an important part of decision-making and judgment, there have been a large number of relevant researches related to commodity order forecasting. For example, in a prior art, such as the patent application publication number CN107038492A, it discloses a method for order forecasting using an ARMA model. However, although the ARMA model can extract some time series features from historical data, it still belongs to the traditional mac...

AI Technical Summary

Problems solved by technology

However, although the ARMA model can extract some time series features from historical data, it still belongs to the traditional macro analysis method.

Due to the very limited complexity of the model, the types of features that can be represented are relatively single, resulting in low accuracy of the prediction results of this method, and the scope of application is greatly limited.

[0004] In the prior art, there is also a commodity order prediction method that introduces wavelet neural network. However, since the neural network only has a single hidden layer perceptron, there is also the problem of insufficient feature extraction capability, which leads to the prediction results obtained by this prediction method. Product order evaluation operation is not very helpful

Images