Flow quantity prediction multi-time-sequence model generation method, information sending method and device

Abstract

The embodiment of the invention discloses a flow quantity prediction multi-time-sequence model generation method, an information sending method and an information sending device. A specific embodiment of the method comprises: obtaining a historical flow volume set of a target article in a preset time period; inputting the historical flow quantity set into each flow quantity prediction time sequence model to obtain a predicted flow quantity set; determining each historical flow in the predicted flow set and the historical flow set in each back-test time granularity as an input parameter of a preset linearization objective function to generate a to-be-solved objective function; based on the constraint condition set, solving the to-be-solved target function to obtain a model weight coefficient set; and according to the model weight coefficient set, performing weighted combination processing on the flow prediction time sequence model corresponding to each model weight coefficient in the model weight coefficient set to obtain a flow prediction multi-time sequence model. According to the embodiment, the accuracy, robustness and stability of flow prediction are improved, and the model determination process is simplified.

Description

Technical field [0001] Embodiment of the present disclosure relates to computer technologies, and particularly relates to a multi-series circulation amount prediction model generation method, an information transmission method, apparatus, electronic device and computer readable media. Background technique [0002] Demand Forecast (Demand Forecast) is a supply chain is very important part, and inventory planning, supply chain execution referred to as three defense supply chain. Currently, the amount of flow when generating predictable, typically used are: a plurality of timing or amount of data transfer through historical models and selected to generate the amount of flow for some time. [0003] However, when the amount of flow generated using the above prediction mode, often there is a technical problem: When using a single timing model can not cover multi-periodicity of the time series, resulting in poor prediction accuracy and robustness; using a plurality of when the timing mo...

AI Technical Summary

Problems solved by technology

[0003] However, when using the above method to generate forecasted turnover, there are often the following technical problems: when using a single time series model, it cannot cover the multi-periodic characteristics of the time series, resulting in poor forecasting accuracy and robustness; using multiple For time series models, it is necessary to select each time series model in advance and determine the number of time series models and the weight of each time series model. The stability of the prediction results is poor, and the model determination process is relatively cumbersome

Images