Real-time recommendation method, electronic device and computer readable storage medium

Abstract

The invention discloses a real-time recommendation method. The method comprises the steps of creating a model class through a predetermined programming language, offline training models on a specificcomputing engine by utilizing the model class, and mandating trained models to a real-time computing background; through the real-time computing background, storing user features in a high-speed cacheregion; mapping user identifiers to numbers in specified intervals, and allocating users in different number intervals to different models; and calling the different models to obtain the user features corresponding to the users in the different number intervals from the high-speed cache region, performing user behavior prediction according to the obtained user features, and recommending prediction results to the users in real time. An iterative model can be quickly updated, so that transverse expansion is facilitated.

Description

technical field [0001] The invention relates to the field of computer information technology, in particular to a real-time recommendation method, electronic equipment and a computer-readable storage medium. Background technique [0002] At present, the data analysis tools of most big data engines (such as Spark engine) can only process offline data, and the corresponding real-time processing components (such as spark-streaming components) need the cluster support of big data engines, resulting in high cost and long time-consuming (100ms -2s), which cannot meet the low time-consuming requirements of real-time recommendation. In addition, the recommendation models trained by the current big data engine cannot be updated online in time to generate benefits, and it is not convenient to manage multiple models for A / B testing online at the same time. Therefore, the design of the real-time recommendation method in the prior art is not reasonable enough and urgently needs to be imp...

AI Technical Summary

Problems solved by technology

[0002] At present, the data analysis tools of most big data engines (such as Spark engine) can only process offline data, and the corresponding real-time processing components (such as spark-streaming components) need the cluster support of big data engines, resulting in high cost and long time-consuming (100ms -2s), which cannot meet the low time-consuming requirements of real-time recommendation

In addition, the recommendation model trained by the current big data engine cannot be updated online in time to generate benefits, and it is not convenient to manage multiple models for A / B testing online at the same time

Therefore, the design of the real-time recommendation method in the prior art is not reasonable enough and needs to be improved urgently.

Images