Details

Speaker(s):  Dean Abbott 

Duration: 120 minutes

Track: Analyze & Interpret

This is a 2-hour lab session.
This session describes the six stages of predictive analytics projects according to the CRISP-DM predictive modeling framework: Business Understanding, Data Understanding, Data Preparation, Modeling, Evaluation, and Deployment. Each stage is described from the analyst’s perspective, providing insights into what the science tells us about each stage, and where theory falls short to help us make about how to proceed in building models.
After completion of the lab, participants should be able to load data, perform simple data preparation, and create predictive models from modeling data sets. The data preparation steps will include filling missing values and creating dummy variables. Predictive modeling steps will include sampling, building models with decision trees, logistic regression, and neural networks. Even though we are using only the KNIME software, the principles will apply to any workflow-style predictive analytics software package.