How paullo0106/prophet_anomaly_detection works
Project type and purpose
• This is an open-source Python utility library (with a demo notebook) for time-series anomaly detection built on Facebook’s Prophet. It provides three core helper functions (prophet_fit, get_outliers, prophet_plot) that wrap Prophet’s fitting, forecasting, outlier-flagging, and visualization steps.
• The README describes usage in a Jupyter notebook (example.ipynb), points at the utils.py file for the core logic, and stores sample output images under img/.
File structure
• README.md – overview, usage, links.
• example.ipynb – interactive demo that imports and calls utils.py.
• utils.py – implements the three key functions that invoke fbprophet, pandas, and matplotlib.
• img/ – sample anomaly-detection figures.
There are no frontend/backend splits or microservices here—just a single Python module and a notebook.
README for additional insights
• It shows the core workflow:
a) configure a Prophet model,
b) call prophet_fit to train and forecast,
c) call get_outliers to identify anomalies,
d) call prophet_plot to render the time series and highlight outliers.
• External dependencies: pandas (for DataFrame), fbprophet (the forecasting engine), matplotlib (or plotly) for plotting.