BookmarkSubscribeRSS Feed
0 Likes

Currently, the Java/C++ score code generated by SAS Enterprise Miner can only be deployed to Windows/Linux/MacOS environments.  More and more applications are being hosted on Mobile platforms (iOS, Android), and the ability to score data locally and instantaneously on a mobile device is highly desireable.  Consider offering score-code libraries that can be run on Android and/or iOS.

3 Comments
SASKiwi
PROC Star

What is your use case for mobile-based scoring? Most mobile apps where scoring would be useful, for example banking and finance apps, are totally reliant on back-end servers to deliver the required data and scoring analytics. The scoring needs to happen where there is plenty of processing resources and minimal data latency on the often large amounts of data required. I don't see how that would be practical on mobile devices. 

MikeStockstill
SAS Employee

Thank you.  We are reviewing this suggestion.

apeterson
Calcite | Level 5
We do have a use case where it is practical as the "scoring" of new data is often much less computationally intensive than training, and if the sample count for scoring is low (and in our case it is), then the computational burden in terms of time is very reasonable.
In fact, with recent improvements in Mobile technologies and architectures, machine learning on mobile platforms is expected to be a high-growth emerging market. As support for this thinking, please refer to the following:
https://www.tensorflow.org/mobile/tflite/
Google (and others) are already making this capability available for Mobile developers. It would make a lot of sense for SAS to do the same.