Get useful, real-world insights into using predictive analysis and data mining to solve problems.
Overview
Syllabus
Introduction
- Data mining and predictive analytics
- Introducing the essential elements
- Defining data mining
- Introducing CRISP-DM
- Beginning with a solid first step: Problem definition
- Framing the problem in terms of a micro-decision
- Why every model needs an effective intervention strategy
- Evaluate a project's potential with business metrics and ROI
- Translating business problems into data mining problems
- Understanding data requirements
- Gathering historical data
- Meeting the flat file requirement
- Determining your target variable
- Selecting relevant data
- Hints on effective data integration
- Understanding feature engineering
- Developing your craft
- Skill sets and resources that you'll need
- Compare machine learning and statistics
- Assessing team requirements
- Budgeting sufficient time
- Working with subject matter experts
- Anticipating project challenges
- Addressing missing data
- Addressing organizational resistance
- Addressing models that degrade
- Preparing for the modeling phase tasks
- Searching for optimal solutions
- Seeking surprise results
- Establishing proof that the model works
- Embracing a trial and error approach
- Preparing for the deployment phase
- Using probabilities and propensities
- Understanding meta modeling
- Understanding reproducibility
- Preparing for model deployment
- How to approach project documentation
- CRISP-DM and the laws of data mining
- Understanding CRISP-DM
- Advice for using CRISP-DM
- Understanding the nine laws of data mining
- Understanding the first and second laws
- Understanding the data preparation law
- Understanding the laws about patterns
- Understanding the insight and prediction laws
- Understanding the value law
- Understanding why models change
- Next steps
Taught by
Keith McCormick