I'm a Business Intelligence analyst working almost entirely with SQL Server Management Studio (SSMS) and Tableau to produce BI reports. I'd like to start to incorporate some predictive analytics, and there appears to be a panoply of options. I'd love to have the insights of people who have worked in machine learning/data science on how to proceed. It looks like there's a variety of options
My objective criteria are:
1.) Learn and employ predictive analytics somewhat quickly in a healthcare setting.
A lot of the time my outcome variables are binary (did someone relapse or not), my input variables tend to be categorical (which program were they in?), and it's EHR data for people living with mental illness, so concrete metrics are few and far between (what was their skills assessment score?).
2.) Keep data import/export simple between SSMS and whatever tools I end up with
I need ease of reporting and to minimize the number of software platforms being employed--with too many tools our users get overloaded, so I'd like to use the same reporting platform (Tableau right now).
3.) Have a robust user base for whatever ends up being used
I want to be able to learn and troubleshoot without issue or obsolete/niche software that nobody uses. I'm not a great autodidact so it would be nice if there's good online support/tutorials. Network effects rock!
As such, I've been doing cursory investigation of what Microsoft already has: it appears there's R Client, Microsoft Machine Learning Server (formerly Microsoft R Server), and SQL Server Machine Learning Services (In-Database), and plain old R.
The in-database option looks nice, particularly because I could execute stored procedures from the front end and have predictive data results--even if it's just regression analysis results--come back to the user. However, I'd like to be able to employ the various R libraries (ggplot2 looks amazing) and am not sure if that's doable.
R server looks great as well, but again I'd like to be able to keep the outputs integrated with Tableau so users have one tool. If I can still use Tableau to kick off a stored procedure with input parameters that pass to R server and have the resulting data come back to them, that would be great. I just don't know if the results will be savable, and if so, can they be kept in the database used by SSMS.
I have no idea what R client does; given that I work for a small organization with a 6 person IT department, maybe that's a better option. But probably has its downsides as well.
There's probably a lot of confusion here and I hope you'll forgive my naivete, but there's a lot of information out there. I'm trying to come up with a sufficient and satisfactory solution that's got solid justification for purchase before we invest big in something. I'm leaning with Microsoft Machine Learning Server (formerly R Server), but am very open to dissuasion. Any suggestions?