Skip to content

Robyn is a Marketing Mix Modeling (MMM) code. It can be used to build end-to-end time series regression models and as an econometrics code library. Automated, built for very large data sets, and is suitable for digital and complex consumer behaviour

License

Notifications You must be signed in to change notification settings

teytaud/Robyn

 
 

Repository files navigation

FB NextGen MMM beta R script, v19
2020-04-29

###### How To Start ######

The following files are contained in the zip:

Four .R script files:
- fb_nextgen_mmm_v19.exec.R # you only need this script to execute, it calls the other 3 scripts
- fb_nextgen_mmm_v19.func.R # this contains all major functions and data preparation
- fb_nextgen_mmm_v19.plot.R # this contains all plot functions
- fb_nextgen_mmm_v19.bayes.R # this contains adapted functions from the library parBayesianOptimization

Two .csv files:
- de_simulated_data.csv # this is our simulated data set
- generated_holidays.csv # this contains holidays of all countries from the library prophet

All files should be placed in the same folder

Please make sure you've installed all library specified in fb_nextgen_mmm_v19.exec.R first

Initial setting is running 1000 random search trails with geometric adstockinng. After installing all libraries, if you select all and run in fb_nextgen_mmm_v19.exec.R, the script should run and produce one plot

Script execution details please see comments in scripts

## Join the FB NextGen MMM community ##

###### FB Contact ######

gufeng@fb.com, Gufeng Zhou, Marketing Science Partner
leonelsentana@fb.com, Leonel Sentana, Marketing Science Partner
aprada@fb.com, Antonio Prada, Marketing Science Partner
igorskokan@fb.com, Igor Skokan, Marketing Science Partner
kylescotshank@fb.com, Kyle Scot Shank, Marketing Science Partner 

See the [CONTRIBUTING](CONTRIBUTING.md) file for how to help out.

## License
FB NextGen MMM R script is MIT licensed, as found in the LICENSE file.

About

Robyn is a Marketing Mix Modeling (MMM) code. It can be used to build end-to-end time series regression models and as an econometrics code library. Automated, built for very large data sets, and is suitable for digital and complex consumer behaviour

Resources

License

Code of conduct

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • R 100.0%