Throughout the workshop, the CRISP-DM model will be used to guide participants through the steps of the data mining process, and the attendees themselves will complete the entire data mining process during the workshop by solving simple data mining problems through a staged progression.

In the morning, participants will begin with a database containing multiple tables of information. Participants will each have a networked computer and may choose to work on exercises in pairs or individually. Attendees will determine which business questions will be considered, how they will be addressed using data mining, and how the data will be prepared for data mining. A divide-and-conquer approach will be used to carry out the “data understanding” and “data preprocessing” steps as participants work alone or in pairs to prepare the data for data mining.

In the afternoon, regression, decision tree, and neural network models will be created, and performance assessed. Participants may optimize these models by using advanced algorithm options, and report model performance on held-out data and summaries of key variables used in the models. Data preprocessing will be re-applied if models do not meet performance requirements. The team will determine which model best addresses the business question, and score the model on validation data.

Throughout the day, emphasis will be placed not only on the data mining process from a technical perspective, but also how to interpret, explain and apply results that have been discovered during the process.
 

LEVEL II COURSE PARTICIPANTS with an interest in applying the methods and techniques first-hand as presented and illustrated in the course

DATA MINING PRACTITIONERS who wish to expand their skills and analytical toolbox as well as hone proficiencies in maneuvering elusive data mining obstacles that stand in the way of superior model accuracy

BUSINESS ANALYSTS who must develop and interpret models, communicate the results and make actionable recommendations

FUNCTIONAL ANALYSTS: Customer Relationship Managers, Risk Analysts, Statistical Analysts, Business Forecasters, Inventory Flow Analysts, Direct Marketing Analysts, Medical Diagnostic Analysts, Market Timers, e-commerce System Architects and Web Data Analysts

• Driving the Level II presentation material through team exercises
• Hands-on experience through the data mining process via a staged progression of exercises using application data
• First-hand vendor-neutral exposure to various data mining tools
• Real-world perspective of data preparation for data mining, model optimization and results interpretation
• Cross-learning through team exercise comparisons to reveal what worked, what didn't, and why?

COURSE OUTLINE

MORNING

• Team Meeting #1
  • Introduction
       • Purpose of Data Mining III: Practice
       • CRISP-DM
       • Description of Data Source for Modeling
  • Business Understanding
       • Prioritize Questions to be Addressed
       • Determine Method to Score Results
  • Assign Data Understanding Responsibilities
  • Result: List of Prioritized Business Questions and Corresponding Data
  • Mining Approaches
   
• Breakout Session #1: Data Understanding
  • Summary Statistics
  • Visualization
  • Outlier Analysis
  • Missing Data Analysis
  • Create Mini-Report
   
• Team Meeting #2
  • Data Assessment Summary
  • Assign Data Preprocessing Responsibilities
  • Result: Summaries of Available Modeling Data, and Recommendations for
  Their Use
   
• Breakout Session #2: Data Preprocessing
  Correct Data Problems
  Create Features
  Create Mini-Report
   
• Team Meeting #3
  Data Preprocessing Summary
  Join Data Modified During Breakout Sessions
  Determine Sampling Strategy
  Assign Modeling Responsibilities
  Result: Single Modeling Dataset

AFTERNOON

• Breakout Session #3: Modeling and Evaluation
  • Build Decision Trees, Regression, and Neural Networks
  • Assess Results
  • Rebuild Models, Changing Modeling Parameters
   
• Breakout Session #4: Model Evaluation and Assessment
  • Score Models on Testing Data
  • Rank Variable Importance to Models
  • Create Mini-Report
   
• Team Meeting #4
  • Summarize Modeling Results
  • Assess Which Modeling Techniques Worked and Didn’t Work
  • Determine Needs for More Data Pre-Processing and Modeling
  • Assign Responsibilities
   
• Breakout Session #5: Re-visit Modeling
  • Create Final Models
  • Create Mini-Report
  • Score Models on Validation Data
   
• Team Meeting #5
  • Select Final Model to Use (if any)
       • Explain Reason for Selection
       • Assess Trade-offs Between Models
  • Assign List of Action Items Pending
   

Courses May Be Delivered At Your Site

Call (888) 742-2454 or send an email inquiry to receive a value-based
spreadsheet quotation for training at your site.

Ana Lemmon and Elisabetta Halkard
Office of Special Investigations
US Air Force

Eric Rickard
Information Computing Sciences
SRI International
 

David Cousins
Divisional Scientist
BBN Technologies
 

Roger Chua
Executive Vice President
Sencor Analytics
 

Stephen Pearce
Preventive Medicine
Kaiser Permanente
 

Seating is limited to 18 participants.  Register early!
Proceed to the On-Line Registration Form to reserve your space today.