Turning Data into Insight with IBM Machine Learning for z/OS

Turning Data into Insight with IBM Machine Learning for z/OS

by Samantha Buhler, Guanjun Cai, John Goodyear, Edrian Irizarry, Nora Kissari, Zhuo Ling, Nicholas Marion, Aleksandr Petrov, Junfei Shen, Wanting Wang, He Sheng Yang, Dai Yi, Xavier Yuen, Hao Zhang
Released September 2018
Publisher(s): IBM Redbooks
ISBN: 9780738457130

Read it now on the O’Reilly learning platform with a 10-day free trial.

O’Reilly members get unlimited access to books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.

Start your free trial

Book description

The exponential growth in data over the last decade coupled with a drastic drop in cost of storage has enabled organizations to amass a large amount of data. This vast data becomes the new natural resource that these organizations must tap in to innovate and stay ahead of the competition, and they must do so in a secure environment that protects the data throughout its lifecyle and data access in real time at any time.

When it comes to security, nothing can rival IBM® Z, the multi-workload transactional platform that powers the core business processes of the majority of the Fortune 500 enterprises with unmatched security, availability, reliability, and scalability. With core transactions and data originating on IBM Z, it simply makes sense for analytics to exist and run on the same platform.

For years, some businesses chose to move their sensitive data off IBM Z to platforms that include data lakes, Hadoop, and warehouses for analytics processing. However, the massive growth of digital data, the punishing cost of security exposures as well as the unprecedented demand for instant actionable intelligence from data in real time have convinced them to rethink that decision and, instead, embrace the strategy of data gravity for analytics. At the core of data gravity is the conviction that analytics must exist and run where the data resides. An IBM client eloquently compares this change in analytics strategy to a shift from "moving the ocean to the boat to moving the boat to the ocean," where the boat is the analytics and the ocean is the data.

IBM respects and invests heavily on data gravity because it recognizes the tremendous benefits that data gravity can deliver to you, including reduced cost and minimized security risks. IBM Machine Learning for z/OS® is one of the offerings that decidedly move analytics to Z where your mission-critical data resides. In the inherently secure Z environment, your machine learning scoring services can co-exist with your transactional applications and data, supporting high throughput and minimizing response time while delivering consistent service level agreements (SLAs).

This book introduces Machine Learning for z/OS version 1.1.0 and describes its unique value proposition. It provides step-by-step guidance for you to get started with the program, including best practices for capacity planning, installation and configuration, administration and operation. Through a retail example, the book shows how you can use the versatile and intuitive web user interface to quickly train, build, evaluate, and deploy a model. Most importantly, it examines use cases across industries to illustrate how you can easily turn your massive data into valuable insights with Machine Learning for z/OS.

Table of contents

  1. Front cover
  2. Notices
    1. Trademarks
  3. Preface
    1. Authors
    2. Acknowledgements
    3. Now you can become a published author, too!
    4. Comments welcome
    5. Stay connected to IBM Redbooks
  4. Chapter 1. Overview
    1. 1.1 Challenges of exponential data growth
    2. 1.2 Trends in artificial intelligence and cognitive systems
    3. 1.3 IBM’s approach to artificial intelligence and cognitive systems
    4. 1.4 IBM Machine Learning for z/OS: An enterprise machine learning solution
    5. 1.5 Unmatched capabilities of Machine Learning for z/OS
      1. 1.5.1 Secure IBM Z platform for running machine learning with data in place
      2. 1.5.2 Full lifecycle management of models
      3. 1.5.3 Enterprise grade performance and high availability
      4. 1.5.4 Flexible choice of machine learning languages and scoring engines
      5. 1.5.5 Intuitive self-guided modeling capabilities
      6. 1.5.6 Developer-friendly API interfaces for applications on the Z platform
    6. 1.6 Value proposition of Machine Learning for z/OS
  5. Chapter 2. Planning
    1. 2.1 Product installers
    2. 2.2 Hardware and software requirements
      1. 2.2.1 Prerequisites for z/OS
      2. 2.2.2 Prerequisites for Linux
      3. 2.2.3 Prerequisites for Linux on Z
    3. 2.3 System capacity
      1. 2.3.1 Basic system capacity
      2. 2.3.2 Capacity considerations for training services
      3. 2.3.3 Capacity considerations for scoring services
      4. 2.3.4 Capacity considerations for performance
    4. 2.4 Installation options on z/OS
      1. 2.4.1 Option 1: Training and scoring services on the same LPAR
      2. 2.4.2 Option 2: Training and scoring services on different LPARs
      3. 2.4.3 Option 3: Training services that are on an LPAR and scoring services that are on an LPAR cluster
    5. 2.5 User IDs and permissions
    6. 2.6 Networks, ports, and firewall configuration
      1. 2.6.1 Network requirements
      2. 2.6.2 Ports
    7. 2.7 Firewall configuration
  6. Chapter 3. Installation and customization
    1. 3.1 Installation roadmap
    2. 3.2 Installing and configuring IzODA
      1. 3.2.1 Installing IzODA
      2. 3.2.2 Configuring IzODA
      3. 3.2.3 Verifying IzODA installation and configuration
    3. 3.3 Configuring security
      1. 3.3.1 Authorizing and authenticating users
      2. 3.3.2 Creating and distributing an SSL certificate
      3. 3.3.3 Configuring LDAP with SDBM for user authentication
    4. 3.4 Installing and configuring Machine Learning for z/OS
      1. 3.4.1 Installing machine learning services on z/OS
      2. 3.4.2 Installing the scoring service in a CICS region
      3. 3.4.3 Installing machine learning services on Linux or Linux on Z
    5. 3.5 Configuring high availability and scalability
      1. 3.5.1 Configuring TCP/IP for port sharing and load balancing
      2. 3.5.2 Scaling an application cluster with extra compute nodes
  7. Chapter 4. Administration and operation
    1. 4.1 Administration dashboard
      1. 4.1.1 Accessing the administration dashboard
      2. 4.1.2 Dashboard page
      3. 4.1.3 Control Nodes page
      4. 4.1.4 Services page
      5. 4.1.5 Pods page
      6. 4.1.6 Cluster Logs page
      7. 4.1.7 Users page
      8. 4.1.8 System Configuration page
      9. 4.1.9 Scoring Services page
      10. 4.1.10 Spark Clusters page
    2. 4.2 Administering by using the administration dashboard
      1. 4.2.1 Managing users and privileges
      2. 4.2.2 Managing scoring services
      3. 4.2.3 Managing remote Spark clusters
      4. 4.2.4 Adding compute nodes
      5. 4.2.5 Updating system configurations
    3. 4.3 Administering by using commands on Linux or Linux on Z
      1. 4.3.1 Monitoring system resource usage
      2. 4.3.2 Checking the status of the GlusterFS file system
      3. 4.3.3 Checking the status of machine learning services
      4. 4.3.4 Stopping and starting nodes
    4. 4.4 Administering by using commands on z/OS
      1. 4.4.1 Stopping and starting a Spark cluster
      2. 4.4.2 Checking Spark Java processes
      3. 4.4.3 Managing the Jupyter kernel gateway
      4. 4.4.4 Managing stand-alone scoring servers
      5. 4.4.5 Managing a scoring server in a CICS region
  8. Chapter 5. Model development and deployment: A retail example
    1. 5.1 Machine Learning for z/OS web UI
      1. 5.1.1 Signing in the Machine Learning for z/OS web UI
      2. 5.1.2 Community
      3. 5.1.3 Projects
      4. 5.1.4 Model Management
      5. 5.1.5 Environments
    2. 5.2 Machine learning workflow for model development and deployment
      1. 5.2.1 Data collection
      2. 5.2.2 Data preparation
      3. 5.2.3 Model training
      4. 5.2.4 Model evaluation
      5. 5.2.5 Model deployment
    3. 5.3 Developing and deploying a model to predict tent sales
      1. 5.3.1 Creating a project
      2. 5.3.2 Adding a data set
      3. 5.3.3 Training and saving a model
      4. 5.3.4 Publishing a model
      5. 5.3.5 Deploying a model
    4. 5.4 Preparing a model for online scoring by using CICS program ALNSCORE
  9. Chapter 6. Use cases: Applying Machine Learning for z/OS in business
    1. 6.1 Customer churn: Reducing customer attrition in banking
      1. 6.1.1 Analyzing historical churn data
      2. 6.1.2 Building and deploying a churn model
      3. 6.1.3 Monitoring and reevaluating the churn model
    2. 6.2 Investment advisory: Helping clients make the right decisions
      1. 6.2.1 Analyze historical client affinity data
      2. 6.2.2 Defining a pipeline for an affinity model
      3. 6.2.3 Training the affinity model
      4. 6.2.4 Evaluating the affinity model
      5. 6.2.5 Saving the predictions of the affinity model
    3. 6.3 Loan approval: Analyzing credit risk and minimizing loan defaults
      1. 6.3.1 Analyzing historical loan approval data
      2. 6.3.2 Defining a pipeline for the loan approval model
      3. 6.3.3 Training the loan approval model
      4. 6.3.4 Evaluating and testing the loan approval model
    4. 6.4 Fraud detection: Rooting out frauds in government benefit programs
      1. 6.4.1 Overview of historical SNAP data and the SNAP model
      2. 6.4.2 Importing, deploying, and testing the SNAP model
    5. 6.5 ITOA: Detecting system anomalies and resolving issues before they arise
      1. 6.5.1 Preparing historical health tree data
      2. 6.5.2 Building the health tree model
      3. 6.5.3 Building the health tree based on the health tree model
      4. 6.5.4 Monitoring the health of your Db2 subsystem
  10. Appendix A. Machine learning basics
    1. Data collection
    2. Data cleaning
      1. Data exploration
    3. Feature preprocessing
    4. Model building
    5. Model evaluation
  11. Appendix B. R and RStudio
    1. Overview
    2. Why R?
    3. Readability of R
  12. Back cover

Product information

  • Title: Turning Data into Insight with IBM Machine Learning for z/OS
  • Author(s): Samantha Buhler, Guanjun Cai, John Goodyear, Edrian Irizarry, Nora Kissari, Zhuo Ling, Nicholas Marion, Aleksandr Petrov, Junfei Shen, Wanting Wang, He Sheng Yang, Dai Yi, Xavier Yuen, Hao Zhang
  • Release date: September 2018
  • Publisher(s): IBM Redbooks
  • ISBN: 9780738457130