Electronic device and model updating method

1. An electronic device, comprising:

a storage circuit for recording a plurality of modules;

a processor for accessing and executing the plurality of modules, wherein the plurality of modules comprises:

a prediction module for inputting a plurality of files to a first model and outputting a predicted result of each of the plurality of files;

an intelligent labeling module for receiving a corrected result for correcting the predicted result of at least one first file in the plurality of files, and generating a first label file corresponding to the first file according to the corrected result and the first file;

an automatic training module for training a plurality of models according to the first label file to generate a plurality of trained models;

a model evolution decision module for testing the plurality of trained models by using at least one test set; and

an automatic process management and control module for replacing the first model with a first trained model when the predicting accuracy of the first trained model in the plurality of trained models is higher than the predicting accuracy of the first model.

2. The electronic device according to claim 1 , wherein

the automatic process management and control module manages and stores the first label file,

the automatic process management and control module determines whether to execute the operation of training the plurality of models according to the first label file,

the automatic process management and control module selects the plurality of models to be trained,

the automatic process management and control module determines a schedule and a time for executing the operation of training the plurality of models according to the first label file, and

the automatic process management and control module determines whether to execute the operation of testing the plurality of trained models by using the test set.

3. The electronic device according to claim 1 , wherein

the intelligent labeling module does not receive another corrected result for correcting the predicted result of at least one second file in the plurality of files, and generates a second label file corresponding to the second file according to the predicted result of the second file, and

the automatic training module trains the plurality of models according to the second label file to generate the plurality of trained models.

4. The electronic device according to claim 1 , wherein in the operation of training the plurality of models according to the first label file to generate the plurality of trained models,

the automatic training module executes a data pre-processing operation and/or a data enhancement operation on the first label file to generate labeled data corresponding to the first label file, and

the automatic training module trains the plurality of models according to the labeled data to generate the plurality of trained models.

5. The electronic device according to claim 4 , wherein in the operation of training the plurality of models according to the labeled data to generate the plurality of trained models,

the automatic training module trains the plurality of models by using a part of the labeled data to generate the plurality of trained models.

6. The electronic device according to claim 5 , wherein in the operation of testing the plurality of trained models by using the test set,

the model evolution decision module tests the plurality of trained models by using another part of the labeled data.

7. The electronic device according to claim 1 , wherein the test set comprises a golden test set, and the golden test set comprises a plurality of verified labeled data.

8. The electronic device according to claim 1 , wherein

the model evolution decision module outputs a ranking of the plurality of models, and the ranking is used for representing the priority of the plurality of models used to replace the first model.

9. The electronic device according to claim 1 , wherein

the intelligent labeling module inputs the corrected result and the first file to a second model, the second model outputs another predicted result corresponding to the first file according to the corrected result and the first file,

the intelligent labeling module receives another corrected result for correcting another predicted result, and generates a third label file corresponding to the first file according to another corrected result and the first file, and

the automatic training module trains another model according to the third label file, the corrected result and the first file.

10. A model updating method for an electronic device, comprising:

inputting a plurality of files to a first model and outputting a predicted result of each of the plurality of files;

receiving a corrected result for correcting the predicted result of at least one first file in the plurality of files, and generating a first label file corresponding to the first file according to the corrected result and the first file;

training a plurality of models according to the first label file to generate a plurality of trained models;

testing the plurality of trained models by using at least one test set; and

replacing the first model with a first trained model when the predicting accuracy of the first trained model in the plurality of trained models is higher than the predicting accuracy of the first model.