Anomaly Detection With Time Series Forecasting

Anomaly detection time series is a very large and complex field. In the past few years, many tech-niques based on data science were designed in order to improve the efficiency of methods developedfor this purpose. In this paper, we introduce Recurrent Neural Networks (RNNs) with LSTM units, ARIMA and Facebook Prophet library for anomaly detection with time series forcasting. Becauseof the difficulty in obtaining labeled anomaly datasets, an unsupervised technique will be experimented. Unsupervised anomaly detection is the process of detecting abnormal points in a given dataset without prior label for training. An anomaly could become normal during the data evolu-tion, therefore it is important to maintain a dynamic system to monitor the abnormality. While LSTMs and ARIMA are powerful methods for time series forecasting the future, the Prophet package works best with time series that have strong seasonal effects and several seasons of historical data. The Prophet is very powerful with missing data and shifts in the trend, and specially handles anomalies well.

Autorentext

Hi! My name is Hoan Dao. I have the Master degree of Computer Science in Ca'foscari University of Venice, Italy. Currently I'm data scientist at VNPT, Vietnam. My research interests are in the fields of Data Analysis, Machine Learning and Computer Vision.