probabilistic_forecasts_generalized_pareto_distribution_ets

A basic Python implementation of Probabilistic forecasts for anomaly detection as proposed by Rob J Hyndman - 3 July 2024

https://robjhyndman.com/seminars/isf2024.html International Symposium on Forecasting, Dijon, France

https://raw.githubusercontent.com/robjhyndman/forecast-anomalies-talk/main/forecast_anomalies.pdf

When a forecast is very inaccurate, it is sometimes because a poor forecasting model is used, but it can also occur when an unusual observation occurs. A good forecasting model can be used to identify anomalies. The approach taken is to use a probabilistic forecast, and to compute the “density scores” equal to the negative log likelihood of the observations based on the forecast distributions. The density scores provide a measure of how anomalous each observation is, given the forecast density. A large density score indicates that the observation is unlikely, and so is a potential anomaly. On the other hand, typical values will have low density scores. A Generalized Pareto Distribution is fitted to the largest density scores to estimate the probability of each observation being an anomaly.

This implementation uses statespace ExponentialSmoothing. Although holtwinters ExponentialSmoothing could also be used, a seasonal parameter needs to be defined therefore holtwinters ExponentialSmoothing is not suited to zero knowledge, unsupervised analysis. It detects spike peaks and troughs really, similar to many other algorithms.

See the docstrings - https://earthgecko-skyline.readthedocs.io/en/latest/skyline.custom_algorithms.html#module-custom_algorithms.probabilistic_forecasts_generalized_pareto_distribution_ets

See the custom_algorithm source - https://github.com/earthgecko/skyline/blob/master/skyline/custom_algorithms/probabilistic_forecasts_generalized_pareto_distribution_ets.py