Data from a temperature sensor might trend upward throughout the day if the device is outdoors. Similarly, the internal temperature of an exterior device may be lower in the winter. Not all devices are affected by seasonality but for the ones that are, choosing an algorithm that handles seasonality and trends is important. According to a research paper (Automatic Anomaly Detection in the Cloud Via Statistical Learning) from data scientists at Twitter, Seasonal ESD is a machine learning algorithm that takes seasonality and trends to find anomalies regardless of the seasonality.
For this recipe, we are going to use the city of Chicago lake water purity dataset. We are going to pull in the data file we prepared in the Detecting time series anomalies with Luminol recipe.
