This repository features a R Shiny Web App for training Machine Learning models for Segmentation tasks. This web app was initially implementend as part of my master's thesis and now generalized to be published so that others have the chance of making use of it by adapting it for their purposes, for example.
Quickly explore it!
Using the dummy dataset (dummyData.csv - which has German csv notation) you can explore the app by playing with it here. I tried to make it as easy to follow as possible.
What is it capable of?
- upload csv file including a numeric response variable as well as numeric and/or categorcial predictors
- descriptive data exploration
- 4 ML algorithms to choose from for training
- saving and resuse of trained models
- model performance comparison
- prediction on new data and download of results
- model analysis, e.g. feature importance, decision tree visualization
- ...
Running it locally requires R 3.2.4 to enjoy all features.
Limitations of shinyapps.io deployment:
- Hyperparameter tuning is simplified by setting the tuneLength parameter of the caret package to 1 in order to speed up training for demonstration purposes.
- Bayesian Neural Network does not work but the other three algorithms do
- The deployed version above won't allow the installation of the RGtk2 package so that the visualization of the trained decision tree won't work. The RGtk2 package itself is a dependency of the rattle package, which is currently commented out for server deployment. You can include it for local installation.
- It seems as if trained models cannot be saved on the server so that the Criteria Importance feature and the Prediction Tab features will not work