Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Create fit() method for anomaly detection task. #70

Open
v1docq opened this issue Apr 24, 2023 · 2 comments
Open

Create fit() method for anomaly detection task. #70

v1docq opened this issue Apr 24, 2023 · 2 comments
Assignees
Labels
enhancement New feature or request

Comments

@v1docq
Copy link
Collaborator

v1docq commented Apr 24, 2023

image
For anomaly detection in Fedot.Industrial we need 2 strategy. One - with target (traditional classification approach). Second - without target on train, but with target on test. In second strategy we also have 2 scenario. First - using Fedot regression models inside KalmanFilter realisation. To create a target on train test we can use get_x_y_pairs method . In this case our anomaly is a max difference between predicted and actual value. First max diff is beginning and second max diff is the end of 1 anomaly. Second strategy - using simple "online" detectors. In this case we only tuning hyperparms of methods, such as length of moving window and type of statistic? using for anomaly detection (mean, std, max,...). In this case target on train is not obvious.

@v1docq v1docq added the enhancement New feature or request label Apr 24, 2023
@v1docq
Copy link
Collaborator Author

v1docq commented Apr 24, 2023

Industrial+Bamt.pdf
image

@v1docq
Copy link
Collaborator Author

v1docq commented May 2, 2023

image
Подход с байесовской сетью может быть использован при правильной подборе ширины окна. Может быть поставлена как регрессии где целевая метрика RMSE(MAE) а настраиваемый параметр - ширина окна. После того как было получено значение оптимальной ширины окна происходит реализация стратегии внесения аномалий в тренировочный набор данных.

@v1docq v1docq assigned leostre and PvtKaefsky and unassigned valer1435 and technocreep Apr 23, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

When branches are created from issues, their pull requests are automatically linked.

5 participants