This project contains notebooks which have been created to study, implement and analyze various techniques of anomaly detection. Two types of data were studied and analysed, categorical and time series data.
- For time series data, methods studied and implemented were:
- Isolation forest
- Local Outlier Factor
- Autoencoders
- AutoRegressive Integrated Moving Average(ARIMA)
- Moving Average
- Z-score analysis.
- For categorical data an ensemble of multiple classifiers were used:
- Logistic Regression, K-Nearest Neighbor, Support Vector Classifier, Decision Tree Classifier
- Different voting classifiers were analyzed (hard and soft voting classifier).
- Calibration techniques were applied namely: Platt Scaling and Isotonic Regression.