Recommended path tracking method, device and system

Abstract

The invention discloses a recommended path tracking method, system and device. The method comprises the following steps: recording, by a recommendation model, whether each of N tracking points is passed or not in a process of generating a recommended path of a certain recommendation result, wherein N is an integer more than 2; and acquiring a path tacking array used to represent whether each of the N tracking points is passed or not, wherein the path tracking array is used to restore the recommended path for generating the recommendation result. According to the recommended path tracking method, the storage of paths in the recommendation process is greatly reduced by bit storage and a preferred dictionary compression method; and furthermore, the size of stored data is limited, so that local storage can be used instead of remote transmission storage, thereby substantially lowering the requirement of network data transmission.

Description

technical field [0001] The present invention relates to Internet technology, in particular to a recommended path tracking method, device and system. Background technique [0002] In the recommendation system, there are various recommendation strategies, and each strategy will go through various paths of the recommendation model in the recommendation process. How to retain the intermediate data of each recommendation strategy in the online system is a problem. huge challenge. Usually, the more traditional approach is to print logs and recommendation IDs at each point that needs to be recorded during the recommendation process, and then start a separate service to collect these logs and recommendation IDs, and upload these logs and recommendation IDs to a dedicated server for searching and retrieval. [0003] There are many problems with this collection strategy. First of all, since there are a lot of points to be recorded in the recommendation process (thousands for each r...

AI Technical Summary

Problems solved by technology

First of all, since there are a lot of points to be recorded in the recommendation process (thousands of orders per recommendation), the storage space required for each recommendation is relatively large

Secondly, because the generated logs need to start the service for collection, it will occupy the processing capacity of the local server. Thirdly, the frequent upload of logs will lead to a particularly large amount of disk I / O and network I / O

During periods of heavy recommendation tasks, the storage of tracking information will occupy a large amount of system resources

Images