ANOMALY DETECTION SYSTEM APPLYING AN EFFICIENT LEARNING METHOD USING PRIOR KNOWLEDGE AND INTELLIGENT DATA MANAGEMENT

The present invention relates to an anomaly detection system employing an efficient learning method using prior knowledge and intelligent data management. In accordance with the present invention, the anomaly detection system includes: a learning data management subsystem labelling and storing data...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: LEE YONG JIN, JUN JONG ARM, KIM MAL HEE, PYO CHEOL SIG
Format: Patent
Sprache:eng ; kor
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The present invention relates to an anomaly detection system employing an efficient learning method using prior knowledge and intelligent data management. In accordance with the present invention, the anomaly detection system includes: a learning data management subsystem labelling and storing data to be learned (learning data), and classifying and managing the data as previous data and new data; a learning subsystem training a model in at least one of a transfer training method for retraining a part of a trained model and a retraining method for retraining the entire trained model, by using the stored learning data, and then storing the trained model; and an inference subsystem executing the trained model to determine whether data introduced from the outside is normal. Therefore, the present invention is capable of effectively applying an artificial intelligence service to an actual field. 본 발명은 선지식과 지능적 데이터 관리를 활용한 효율적 학습 방법을 적용한 이상감지 시스템에 관한 것이다. 본 발명에 따른 이상감지 시스템은 학습할 데이터(학습 데이터)에 레이블을 부여하여 저장하고, 이전 데이터와 신규 데이터로 구분하여 관리하는 학습데이터 관리 서브시스템; 상기 저장된 학습 데이터를 이용하여, 학습된 모델의 일부만 다시 학습하는 전이학습 및 학습된 모델 전부를 다시 학습하는 재학습 중 적어도 하나의 방식으로 모델을 학습시키고, 상기 학습된 모델을 저장하는 학습 서브시스템; 및 상기 학습된 모델을 실행하여 외부에서 유입된 데이터의 정상 여부를 판단하는 추론 서브시스템;을 포함한다.