Prophet - Time Series Forecasting from Meta (facebook)

Prophet - Time Series Forecasting #

Time series forecasting is a common problem in every corner of our work life. We usually use excel to generate the forecast if the data is linear. However many scenarios, we do not get accurate results.

Welcome to Prophet a project from Meta (facebook). Prophet provides time series model which is extensible enough for a wide range of business time series, yet configurable by non-experts who may have domain knowledge about the data generating process but little knowledge about time series models and methods.

Why Prophet? #

• Accurate and fast - produces accurate forecasts, built on top of stan

• Fully automatic - provides reasonable forecast on messy data with no manual effort.

• Tunable forecasts - easy to tune your forecasts.

• Available in both Python and R

White Paper #

Read the white paper at https://peerj.com/preprints/3190/.

Use Case I tried to solve #

The use case I have tried to solve is to predict the share value of Amazon. The time series data is available from 2007 to 2017 and I am predicting the value of the share prices for next 120 days.

Here are some sample predictions prophet made.

Source Code #

Please refer to https://github.com/vensr/prophet-time-series-forecast for the source code and detailed documentation on how to use this tool.

References #

• https://facebook.github.io/prophet/