Industrial raw material consumption prediction method based on multi-task time sequence learning

Abstract

The invention provides an industrial raw material consumption prediction method based on multi-task time sequence learning, and the method comprises the steps: firstly obtaining a historical consumption time sequence set of all raw materials as the input of a model, carrying out the clustering of an original single time sequence set, and further dividing the raw materials into different groups with positive correlation or negative correlation, and serving as a priori constraint of the prediction model. Then, based on a multi-task learning thought, an auxiliary task for predicting long-term and recent future development trends is constructed, space and time dimension features of a time sequence are fully extracted for all tasks based on a prediction model of a neural network, and data features learned among different tasks are shared in the process; the main task is helped to fuse more time sequence information, and the consumption of various raw materials in a period of time in the future is predicted by combining an autoregression model based on a combined model thought.

Description

technical field [0001] The present invention relates to the field of time series prediction, and more specifically, relates to a method for predicting industrial raw material consumption based on multi-task time series learning. Background technique [0002] In actual industrial production, due to inaccurate forecasting of raw material consumption, there may be excess inventory or insufficient raw materials in production. In order to ensure normal production, enterprises usually stock up in excess, but this will increase costs. Accurate raw material demand forecasting can help enterprises determine procurement plans and formulate production plans, and then provide scientific and effective decision support for managing raw material inventory, and realize appropriate inventory to save enterprise costs. Therefore, it is of great economic value to predict the consumption of raw materials for industrial production in the future. However, due to the influence of various internal ...

AI Technical Summary

Problems solved by technology

Due to the neglect of the correlation between the consumption of different types of raw materials and the lack of an effective mechanism to capture its correlation law, the prediction accuracy of traditional methods is usually not satisfactory, and cannot meet the high-precision and low-cost demand for raw material consumption prediction in industry

According to research, there is currently a lack of effective methods to accurately predict raw material consumption

Images