Data organization form representing cash flow and prediction method based on multi-task learning

Abstract

The invention relates to big data processing technology, and aims to provide a data organization form representing cash flow and a prediction method based on multi-task learning. Including: information mining and statistical analysis of historical data of sales flow and electricity consumption in the power sector; establishment of multiple tasks related to regression analysis, establishment of multi-dimensional data labels; cross-validation according to time series, using deep convolutional neural network or The recurrent neural network performs multi-task learning and performs performance testing on the model; the optimal hyperparameters of the neural network are obtained by using the grid method, and finally the configuration of the neural network model is determined, and the neural network model is used to predict the amount of electricity sales. The invention constructs a new data organization form combining these information, which can describe the source of daily cash flow. Compared with the traditional statistical model, the multi-task learning constructed by the present invention requires less manual intervention, and the result is more robust and more adaptable to big data.

Description

technical field [0001] The invention relates to big data processing, in particular to a data organization form representing cash flow and a prediction method based on multi-task learning. Background technique [0002] Sales amount forecast refers to the estimation of the sales quantity and sales amount of all products or specific products within a certain period of time in the future. Sales forecasting aims to put forward feasible sales targets through certain analysis methods on the basis of fully considering various influencing factors in the future, and to help enterprises make financial budgets. The results are of great significance to the development planning and strategic deployment of enterprises . [0003] Still, producing high-quality consumption forecasts is no easy task. The data mining tools currently available for cash flow forecasting are mainly some statistical analysis methods, such as time series analysis, linear / nonlinear regression model, gray system mod...

AI Technical Summary

Problems solved by technology

The traditional method either selects a small number of data dimensions, or ignores the overall correlation between the existing data, which leads to the lack of data information, making further modeling limited a priori; on the other hand, based on all available data Data dimensionality, traditional methods cannot effectively extract high-dimensional, interactive and effective features that are conducive to prediction

However, there have been no reports on the application of neural networks and multi-task learning in grid sales forecasting

Images