"Official" January 2020 BARUG Meeting

Our January BARUG meetup will be held in the evening after the first day of workshops at rstudio::conf, onsite at the conference venue. The meeting is free for all BARUG members and does not require a conference ticket. The BARUG meeting is not officially part of the conference, but we expect several RStudio team members to drop by.

Agenda
6:30 - Food and networking
7:00 - Announcements
7:05 - Sydeaka Watson - Neural Networks for Longitudinal Data Analysis
7:35 - Bryan Lewis - Survival analysis, Salesforce, Shiny, and Shinyproxy
8:05 - Max Kuhn - Creating a modeling package in 9 easy steps

###########
Sydeaka Watson
Dr Watson will give a preview of her rstudio::conf talk:
Neural Networks for Longitudinal Data Analysis
Longitudinal data (or panel data) arise when observations are recorded on the same individuals at multiple points in time. For example, a longitudinal baseball study might track individual player characteristics (team affiliation, age, height, weight, etc.) and outcomes (batting average, stolen bases, runs, strikeouts, etc.) over multiple seasons, where the number of seasons could vary across players. Neural network frameworks such as Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRUs) can flexibly accommodate this data structure while preserving and exploiting temporal relationships. In this presentation, we highlight the use of neural networks for longitudinal data analysis with tensorflow and keras in R.

##########
Bryan Lewis
Survival analysis, Salesforce, Shiny, and Shinyproxy
Part I
A sales and marketing pipeline model tries to predict sales from a time series of observations associated with given accounts. The observations may have many features, importantly including labels of reasonably well-defined stages of the sales process. Salesfore conveniently exports such observations in easy to analyze form. Viewed through the right kind of lens, models like this look a bit like classical survival models. A friend of mine and I build some nice
models along these lines. The first part of my talk briefly discusses such models.

Part II
In order to make the models easy to use, we designed a shiny app. In order to make the app available to our clients, at least in beta form, we deployed it with Shinyproxy (and also Docker and Nginx), a tool that makes deploying Shiny apps pretty easy to do. I'll discuss this delightful deployment process in this 2nd part.

##########
Max Kuhn
I'll create that modeling package in 9 easy steps!
Creating an R package can seem daunting. The {usethis} package has made this process much easier but, for modeling packages, the {hardhat} package can do even more. If you are new to R or experienced, {hardhat} can automatically create the interface code (i.e., formula, non-formula, ...) and populate S3 methods (such as predict) for you. I'll demonstrate this for a truly awful model in a live coding exercise.