Information pushing method and device for small-flow experiment system

Abstract

The embodiment of the invention provides an information pushing method and device for a small-flow experiment system, and belongs to the technical field of testing. The method comprises the followingsteps: selecting a preset proportion of users; and responding to the received request of the user about the information flow content, if the information flow content pushed to the user at the previoustime is generated by a baseline strategy; if the information flow content pushed to the user last time is generated by the experiment strategy, pushing the first information flow content generated bythe experiment strategy to the user at the current time, and if the information flow content pushed to the user last time is generated by the experiment strategy, pushing the second information flowcontent generated by the baseline strategy to the user at the current time. The sensitivity of a small-flow experiment system can be improved, and the time cost is remarkably reduced.

Description

technical field [0001] The invention relates to the technical field of testing, in particular to an information push method and device for a small flow experimental system. Background technique [0002] In information flow products, the evaluation method refers to the method of selecting the best among multiple iterative schemes (or strategies) during the product iterative optimization process. [0003] In Internet information flow products, the information presented by each user is personalized. Under such product characteristics, in order to improve user experience and product effects, various strategies are often introduced to improve the degree of personalization of information flow products, and such optimization adjustments are very frequent. In order to judge whether the adjusted strategy has a positive impact, a small traffic experiment system is usually established. The small traffic experiment system refers to the system that selects some users (or traffic) inste...

AI Technical Summary

Problems solved by technology

[0007] (2) In order to eliminate crowd bias, a certain amount of time is required to idle, and the existence of the idle period increases the overall time cost of the experiment

[0008] (3) The sensitivity of the experimental system is low, that is, the strategy needs to change a lot before the experimental system can judge whether there is a positive return

This solution does not consider the user experience of the information flow. When interweaving the content generated by the experimental strategy and the baseline strategy by item, the diversity of the content is often not guaranteed, and the diversity experience is inconsistent after each refresh.

Images