Anomaly detection in set-top boxes using machine learning technologies
- MSC DA 2017-2019
Anomaly detection is a common goal shared by dierent domains. Anomalies are generally dened as an error or as an unexpected event in reality .It is the practice of identifying items or events that do not conform to an expected behavior or do not correlate with other items in a dataset. Anomalies are detected in performance data of set top boxes using both supervised and unsupervised machine learning Techniques.The eectiveness of each algorithm was evaluated and compared.Which have shown sucient performance and sensitivity. Anomalies are patterns in data that do not conform to a welldened notion of normal behavior.Anomaly detection is the process of identifying unexpected items or events in datasets,which dier from the norm. In contrast to standard classication tasks, anomaly detection is often applied on unlabeled data, taking only the internal structure of the dataset into account. This challenge is known as unsupervised anomaly detection and is addressed in many practical applications.This project Principal Component Analysis (PCA) for feature selection, then unsupervised machinelearning techniques like K-Means are applied on the detected anomalies to dene their classes.supervised machine learning techniques like KNN,OCSVM and Isolation forest. The results show that OCSVM mostly give better results than Isolation forest , but for certain anomalies KNN give the best results