Abstract:
|
In any analytics project, preparing data for modeling can be time-consuming and tedious. In this talk we use Colorado real estate data to illustrate the connection between data visualization, data preparation, and analytics. We use linked and interactive graphics to explore data in many dimensions, and use insights gained from the visual exploration to clean the data and to create new features for modeling. We fit and explore a variety of predictive models (regression, trees, neural networks, and penalized regression), and use graphical tools to compare competing models. Finally, we use interactive visualizations to interpret the final model and summarize model performance.
|