Model management system and method dealing with multiple-business scenarios

Abstract

The invention discloses a model management system and method dealing with multiple-business scenarios. The model management system comprises a model conversion module, a model uploading module, a model pre-compiling module, a model test module and a model management module; the model conversion module converts offline model files into model files with predetermined formats; the model uploading module uploads the model files and loading the same to cache; the model pre-compiling module pre-compiles the uploaded model files; the model test module performs production environment test on the model files, if succeeds, the model files are uploaded, and if fails, the model files are deleted from the cache; the model management module receives requests of clients, acquire the corresponding model files as required, and processes the requests by executing the model files. The process of models from delivery to uploading is unified, risk controllable is achieved, unified management of the models is facilitated, and the situation about inconsistent model results caused by environment different is effectively prevented.

Description

technical field [0001] The invention relates to a model management system and method for multi-service scenarios. Background technique [0002] At present, there are the following defects in model management in multi-business scenarios: [0003] 1. The data analysis team will establish various prediction models for various businesses, and the performance of these models varies due to factors such as modeling language and modeling methods, which makes the entire process from delivery to online of the model uncontrollable; [0004] 2. The representation of the model is different, it may be a series of adjusted parameters, or it may be a persistent file, and the model files produced by different programming languages ​​are inconsistent, which makes it inconvenient to manage uniformly; [0005] 3. In the past, the model was tested in an offline test environment before going online, but after going online to the production environment, the model results may be inconsistent due t...

AI Technical Summary

Problems solved by technology

[0008] The technical problem to be solved by the present invention is to overcome the defects in the prior art that the whole process of the model from delivery to online is uncontrollable and inconvenient for unified management, and to provide a model management system and method for multi-business scenarios

Images