| Autor: |
T. R., Soumya, Revathy, S. |
| Jazyk: |
angličtina |
| Rok vydání: |
2023 |
| Předmět: |
|
| Zdroj: |
International Journal of Intelligent Systems and Applications in Engineering; Vol. 11 No. 7s (2023): Advancements in Machine Learning for Computer Science and Decision Support Systems; 311-316 |
| ISSN: |
2147-6799 |
| Popis: |
In industrial automation, high-dimensional data streams have become more common. The system's dependability and stability can be ensured if system defects can be detected efficiently using this data. The "curse of dimensionality" and conception wandering are the two fundamental problems induced by increased dimensionality and digital data, and one ongoing goal is to handle them both at the same time. This work proposes a method for detecting faults in non - stationarity higher dimensional streaming data. To discover low-dimensional subdomain defects from high-dimensional datasets, anomaly detection technique with subspace mechanism tends to be presented.It calculates the variance inflation of an entity in its subdomain translation and identifies fault tolerant categories by assessing different angles. The technique is then developed to an online application that can continually monitor model parameters depending on the feature extraction concept. On synthetic datasets, we implemented the accuracy technique to local outlier factor-based ways to evaluate it, and discovered that the technique was more accurate. The research's findings showed the proposed computation effectiveness. Researchers further claimed that the method can detect low impact complexity defects those baseline characteristics in high dimensional complexity and it can adapt to analyzed system's timely behaviour. The major contribution of this research is an operational subspace learning technique for defect identification. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
