Putting Machine Learning to Work

Supervised learning solves modern analytics challenges and drives informed organizational decisions. Although the predictive power of machine learning models can be very impressive, there is no benefit unless they inform value-focused actions. Models must be deployed in an automated fashion to continually support decision making for residual impact. And while unsupervised methods open powerful analytic opportunities, they do not come with a clear path to deployment. This course will clarify when each approach best fits the business need and show you how to derive value from both approaches.

Regression, decision trees, neural networks – along with many other supervised learning techniques – provide powerful predictive insights when historical outcome data is available. Once built, supervised learning models produce a propensity score which can be used to support or automate decision making throughout the organization. We will explore how these moving parts fit together strategically.

Unsupervised methods like cluster analysis, anomaly detection, and association rules are exploratory in nature and don’t generate a propensity score in the same way that supervised learning methods do. So how do you take these models and automate them in support of organizational decision-making? This course will show you how.

This course will demonstrate a variety of examples starting with the exploration and interpretation of candidate models and their applications. Options for acting on results will be explored. You will also observe how a mixture of models including business rules, supervised models, and unsupervised models are used together in real world situations for various problems like insurance and fraud detection.

You Will Learn

• When to apply supervised versus unsupervised modeling methods
• Options for inserting machine learning into the decision making of your organization
• How to use multiple models for estimation and classification
• Effective techniques for deploying the results of unsupervised learning
• Interpret and monitor your models for continual improvement
• How to creatively combine supervised and unsupervised models for greater performance

Who is it for?

Analytic Practitioners, Data Scientists, IT Professionals, Technology Planners, Consultants, Business Analysts, Analytic Project Leaders.

Topics Covered

1. Model Development Introduction

Current Trends in AI, Machine Learning and Predictive Analytics

• Algorithms in the News: Deep Learning
• The Modeling Software Landscape
• The Rise of R and Python: The Impact on Modeling and Deployment
• Do I Need to Know About Statistics to Build Predictive Models?

2. Strategic and Tactical Considerations in Binary Classification

• What’s is an Algorithm?
• Is a “Black Box” Algorithm an Option for Me?
• Issues Unique to Classification Problems
  • Why Classification Projects are So Common
  • Why are there so many Algorithms?

3. Data Preparation for Supervised Models

• Data Preparation Law
• Integrate Data Subtasks
  • Aggregations: Numerous Options
  • Restructure: Numerous Options
  • Data Construction
    • Ratios and Deltas
    • Date Math
    • Extract Subtask

4. The Tasks of the Model Phase

• Optimizing Data for Different Algorithms
• Model Assessment
  • Evaluate Model Results
    • Check Plausibility
    • Check Reliability
  • Model Accuracy and Stability
  • Lift and Gains Charts
• Modeling Demonstration
  • Assess Model Viability
  • Select Final Models
• Why Accuracy and Stability are Not Enough
• What to Look for in Model Performance
• Exercise Breakout Session
  • Select Final Models
  • Create & Document Modeling Plan
  • Determine Readiness for Deployment
• What are Potential Deployment Challenges for Each Candidate Model?

5. What is Unsupervised Learning?

• Clustering
• Association Rules
• Why most organizations utilize unsupervised methods poorly
  • Case Study 1: Finding a new opportunity
  • Case Studies 2, 3, and 4: How do supervised and unsupervised work together
  • Exercise Breakout Session: Pick the right approach for each case study
• Data Preparation for Unsupervised
  • The importance of standardization
  • Running an analysis directly on transactional data
• Unsupervised Algorithms:
  • Hierarchical Clustering
  • K-means
  • Self-Organizing Maps
  • K Nearest Neighbors
  • Association Rules
• Interpreting Unsupervised
  • Exercise Breakout Session: Which value of K is best?
  • Choosing the right level of granularity
  • Reporting unsupervised results

6. Wrap-up and Next Steps

• Supplementary Materials and Resources
• Conferences and Communities
• Get Started on a Project!
• Options for Implementation Oversight and Collaborative Development