Collaborative task prediction method and computer readable storage medium

Abstract

The invention provides a collaborative task prediction method and a computer readable storage medium. The method comprises the following steps: receiving sample identity information sent by a plurality of terminals, and constructing a longitudinal model based on the sample identity information through a child longitudinal model obtained by performing intermediate result interaction with the plurality of terminals; receiving a plurality of child transverse models sent by the plurality of terminals, and integrating the plurality of child transverse models to obtain a transverse model; integrating the longitudinal model and the transverse model to obtain a task prediction model, and sending the task prediction model to a plurality of terminals; obtaining to-be-predicted identity information,and sending the to-be-predicted identity information to the plurality of terminals; when a plurality of sub-prediction results returned by the plurality of terminals for the to-be-predicted identity information are received, integrating a task prediction result by using the plurality of child prediction results to complete collaborative task prediction. According to the invention, the accuracy oftask prediction can be improved.

Description

technical field [0001] The invention relates to artificial intelligence technology, in particular to an artificial intelligence-based collaborative task prediction method and a computer-readable storage medium. Background technique [0002] With the development of artificial intelligence technology, more and more industries use artificial intelligence technology to perform various task predictions, such as the financial industry and social media industry. When using artificial intelligence for task prediction, it is usually first to use sample data to train a prediction model, and then perform task prediction based on the prediction model. For the data with strong privacy in the training data, you can use the federated training mode to complete the training, that is, each participant first calculates the gradient separately, the server integrates the total gradient of each participant, and finally each participant uses the integrated total gradient. Gradient update model. ...

AI Technical Summary

Problems solved by technology

However, misaligned data is very likely to contain useful information for task prediction, and the low utilization rate of misaligned data will lead to low performance of the trained prediction model, thus making the accuracy of task prediction low

Images