/******************************************************************************

Copyright (c) 2015 by SAS Institute Inc., Cary, NC 27513 USA

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

******************************************************************************/

SAS_ML_examples

===============

Code and materials for several machine learning tasks with SAS.

To accompany "An Overview of Machine Learning with SAS Enterprise Miner"

Paper available at:

http://support.sas.com/resources/papers/proceedings14/SAS313-2014.pdf

Be sure to unzip the data!

This repo contains samples. The original larger data sources should be available on kaggle.com.

The claim prediction challenge example is completed using the SAS Enterprise Miner GUI. Both a model package and diagram XML are provided to recreate the workflow.

You can use the -MEMSIZE option to increase the amount of RAM availabe to a SAS session.

These examples were tested in the following environment:

Windows Server 2008 R2 Enterprise

Dual Intel Xeon E5-2667 @ 2.9 GHz

128 GB RAM (Most the examples were also run on a laptop with 16 GB of RAM.)

SAS 9.4 (TS1M2)

SAS Enterprise Miner 13.2

R 2.15.3

===============

emc_israel_data_science_challenge.sas

===============

1.) Download (and unzip) or clone the SAS_ML_examples repository to a directory referred to as {WORK_DIR}.

2.) Unzip example data file (data.zip) so that the following files are extracted to {WORK_DIR}\data.

3.) Open the emc_israel_data_science_challenge.sas file in a standard DMS SAS Session.

4.) Set the git_repo_data_dir macro variable to the (unzipped!) data subdirectory of the repository i.e., {WORK_DIR}\data.

5.) Set the cpu_count macro variable to an integer less than or equal to the number of physical CPU cores on your system.

6.) Submit the entire file or in parts if interested in walking through the steps.

===============

claim_prediction_challenge.xml

===============

This example is in the form of xml that can be imported into SAS Enterprise Miner as a process flow. It also needs R (with cluster package) to be installed alongside SAS.

1.) Install R and set RLANG option:

1a.) Install R (with cluster package)

SAS/IML Version: 13.1 -> Recommended versions of R: 2.13.0-3.0.2 -> Required Version of PMML Package: pmml_1.4.1

SAS/IML Version: 13.2 -> Recommended versions of R: 2.15.3-3.0.3 -> Required Version of PMML Package: pmml_1.4.1

SAS/IML Version: 14.1 -> Recommended versions of R: 3.0.1-3.1.2 -> Required Version of PMML Package: pmml_1.4.2

1b.) If it is not set, set the -RLANG SAS system option:

The –RLANG option is typically added to the main sasv9.cfg file of a given SAS installation.

The main sasv9.cfg file is often located in a directory like

C:\Program Files\SASHome\SASFoundation\9.4\nls\en

Once you have found this file, add

-RLANG

to the bottom of the file.

2.) Download (and unzip) or clone the SAS_ML_examples repository to a directory referred to as {WORK_DIR}.

3.) Unzip example data file (data.zip) so that the following files are extracted to {WORK_DIR}\data.

4.) Create a new project in SAS Enterprise Miner.

5.) In top-left pane, right-click on "Diagrams" folder and select "Import Diagram from XML...". In the Open window, browse to {WORK_DIR} and select claim_prediction_challenge.xml file. This opens the flow diagram.

6.) The first node (CLAIMS_TRAIN_SAMPLE) in the flow represents input data and needs to be re-created to point to the actual data. To do this:

6a.) Create a new library in Enterprise Miner using File->New->Library pointing to {WORK_DIR}\data directory.

6b.) Create a new data source for claims_train_sample.sas7bdat. To do this use File->New->Data Source and the library created in above step to locate the dataset. In addition, select Advanced button for Metadata Advisor Options on Step 4 of the Data Source Wizard creation.

6c.) After the data source is created, validate its metadata is accurate. For that, on top-left pane, under Data Sources, right-click CLAIMS_TRAIN_SAMPLE and select "Edit Variables".

IMPORTANT: Make sure the table metadata matches as below. If not, make necessary changes.

Name Role Level

======== ======== ========

Blind_Make Rejected Nominal

Blind_Model Rejected Nominal

Blind_Submodel Rejected Nominal

Calendar_Year Input Nominal

Cat1 Input Nominal

Cat10 Input Nominal

Cat11 Input Nominal

Cat12 Input Nominal

Cat2 Input Nominal

Cat3 Input Nominal

Cat4 Input Nominal

Cat5 Input Nominal

Cat6 Input Nominal

Cat7 Input Nominal

Cat8 Input Nominal

Cat9 Input Binary

Household_ID ID Nominal

Model_Year Input Interval

NVCat Input Nominal

NVVar1 Input Interval

NVVar2 Input Interval

NVVar3 Input Interval

NVVar4 Input Interval

OrdCat Input Ordinal

Row_ID ID Interval

Var1 Input Interval

Var2 Input Interval

Var3 Input Interval

Var4 Input Interval

Var5 Input Interval

Var6 Input Interval

Var7 Input Interval

Var8 Input Interval

Vehicle Input Nominal

dataobs ID Interval

6d.) Delete existing CLAIMS_TRAIN_SAMPLE node in the flow and drag-drop the newly created data source with same name on to the diagram and connect it to the second node (Variable Selection) in the flow.

7.) Run the flow.

NOTE: The SAS program file claim_prediction_challenge.sas and the model package file claim_prediction_challenge.spk are provided for redunancy.

===============

digit_recognizer.sas

===============

1.) Download (and unzip) or clone the SAS_ML_examples repository to a directory referred to as {WORK_DIR}.

2.) Unzip example data file (data.zip) so that the following files are extracted to {WORK_DIR}\data

3.) Open the digit_recognizer.sas file in a standard DMS SAS Session.

4.) Set the git_repo_data_dir macro variable to the (unzipped!) data subdirectory of the repository i.e., {WORK_DIR}\data.

5.) Set the cpu_count macro variable to an integer less than or equal to the number of physical CPU cores on your system.

6.) Submit the entire file or in parts if interested in walking through the steps.