Recommendation system recall method and system based on parameter configuration

Abstract

The invention discloses a recommendation system recall method and a system based on parameter configuration. A fusion strategy capable of parameter configuration is used; the real-time interest tendency of the user is considered in a recall result; the offline long-term hobby requirements of the user are also considered; therefore, interests and hobbies of the user are truly reflected; automatic fusion of different Batch recall results is realized; considering the parameter weight based on the service understanding into the recall set fusion; parameter change, the recall set generation flexibility is greatly improved, abtest test comparison between different recall strategies is facilitated, an algorithm engineer can be helped to find a good balance point between short-term interests and long-term interests of the recall set, meanwhile, the diversity of subsequent fine sorting results is guaranteed, recommendation results are more reasonable, and the user experience is improved.

Description

Technical field [0001] The invention relates to the technical field of data processing, in particular to a method and system for recalling a recommendation system based on parameter configuration. Background technique [0002] In the online music platform, the audio album recommendation system usually recommends the audio albums that users are interested in. The most important part of the audio album recommendation system is the recommendation algorithm. The recommendation algorithm includes coarse ranking recall and fine ranking ranking parts. According to the different business characteristics, it is necessary to comprehensively consider the relationship between the coarse and fine recall. In the recall generation part, the user's long-term behavior, that is, the historical behavior, needs to be considered, as well as the user's real-time behavior needs. Both have their own advantages and disadvantages. The historical behavior update time is slow, but the data can be considere...

AI Technical Summary

Problems solved by technology

Both have their own advantages and disadvantages. The historical behavior update time is slow, but the data can be considered from the overall situation. The real-time behavior update speed is fast, which can quickly reflect the real-time interest preferences of users, but the data is relatively small.

Images