Noob question: SAP Predictive Analytics vs SAP Hana + R

[Former Member]

Hello SAP world!

i'm a newbie to the SAP world so excuse me if the question looks dumb

So, as far as i can understand i can create mostly any kind of model using my R code integrated on SAP Hana.

So, why would i want to use SAP Predictive Analytics?

Do i miss smth?

Thanks in advance to everyone!

igor.

[PPaolo · Accepted solution]

Hello Igor,

    that's actually a good question!

Using R in HANA is a very good approach but you have to be a good developer and know how to use HANA Studio on top of being expert in predictive algorithms.

If you want to think more about the pure 'data science' part of your tasks rather than on the 'IT' part of development then a tool such as SAP Predictive Analytics can help.

With SAP Predictive Analytics e.g. you have a UI for doing data preparation instead of developing code for that. Similarly, in the application you have pre-build visualizations which can show you the results of your algorithms and help in their analysis.

Moreover SAP Predictive Analytics already provides ready-to-use algorithms, can connect to the existing HANA predictive libraries or can consume R: you just have to drag & drop one of those algorythms and not worry about writing them.

Last but not least, SAP Predictive Analytics contains an 'Automated' mode where various tasks which usually take a lot of time are automated as much as possible (e.g. a large part of data preparation, the application of the model to new data, the deployment of the model into a database or inside an existing programming code and the lifecycle management of your models).

This is quite a short summary of what you can do with SAP Predictive Analytics, overall I would say that the value is to let you concentrate on the high value tasks of a predictive project by taking care of some 'routine' tasks.

Hope that it helps

Best regards

PPaolo

[Former Member · Accepted solution]

, thanks for your answer!

meanwhile, i also watched a couple of training videos and it looks like R-based solution in Hana could be slower than 'native' APL/PAL functions due to data transfer from Hana server instance to R-server instance and back.

So, i guess for people that come from 'R world' it might be an option to use R during 'development' stage and then re-implement the ultimate solution using native APL/PAL functions for 'production' stage. Would you agree on that?

Thanks!

igor.

[achab · Accepted solution]

Hi Igor,

From my perspective what Pierpaolo also wants to point out is that you do not necessarily to be a developer to automatically create performing predictive models on top of SAP HANA data.

This was the value proposition of SAP Predictive Analytics is about, being able to manage end-to-end your data mining project without necessarily doing development, and benefiting at the same time from powerful automated algorithms.

Thanks & regards

Antoine

[PPaolo · Accepted solution]

Hello Igor,

as Antoine pointed out my original answer was on the positioning of SAP Predictive Analytics as an alternative to a fully and manually developed predictive project.

As per your latest question:

So, i guess for people that come from 'R world' it might be an option to use R during 'development' stage and then re-implement the ultimate solution using native APL/PAL functions for 'production' stage. Would you agree on that?

Well, I don't fully agree: why would you want to develop in R and then re-develop the same functionality using the HANA libraries?  I agree that R can be a good way to learn how to build and deal with predictive algorithms (and reuse what you already have built) but once you are confident with that I think that you might want to start developing directly with PAL or APL if R doesnt's statsify your performance requirements.  Moreover there is no guarantee that the results you get with one language will exactly be the same with another language.

As a side note, if you prefer to work in R, please notice that SAP Predictive Analytics 2.3 can execute R algorithms both locally on your workstation (no need to have an HANA server) or from the HANA server itself. In both ways, you define your R function and then use it in your project as a simple drag&drop components of the workflow.

Best regards

PPaolo

[achab]

Hi Igor,

Did we answer your questions? If such is the case, please flag the thread as "Answered".

Thanks for your help,

Antoine