[Forum SIS] Tomorrow: Webinar, 3 Ways to Improve your Regression with Data Science and Machine Learnging (no charge, part 2)

[Lisa Solomon]

3 Ways to Improve your Regression with Data Science and Machine Learning Part 2
 (no charge, Case Study, Step-by-step, Hands-on option)


Registration: http://hubs.ly/H01Y6dj0

 Alternative Link: http://info.salford-systems.com/3-ways-to-improve-your-regression-part2


January 27th, 10AM - 11AM PT

*         If the time is inconvenient, please register and we will send you a recording

*         Part 1 is not required, to understand the approach and concepts in tomorrow's webinar; but, if you want a refresher, you can see last week's webinar at your convenience.  Link to recording of Part 1: http://hubs.ly/H01T4kR0
ABSTRACT:
Last week, we showed you how you could drastically improve prediction accuracy in your linear  regression with a new model that handles missing values, interactions, AND nonlinearities in your data.  As a follow-up to the last week's webinar, we will show you how to take data science techniques even further to extract actionable insight and take advantage of advanced modeling features. You will walk away with several different methods to turn your ordinary regression into an extraordinary regression!

Techniques used:

*         Stochastic gradient boosting: TreeNet plots show you the impact of every variable in your model; take it a step further by creating spline approximations to these variables and using them in a conventional linear regression for a boosted model performance!

*         Nonlinear regression splines: MARS nonlinear regression will still give you what looks like a standard regression equation, but instead of coefficients, you'll see transformations of your original variables.

*         Modeling automation: learn how to cycle through numerous modeling scenarios automatically to discover best-fit parameters.
Included with Registration:

*         On-demand recording of webinar

*         Data set used in presentation

*         Step-by-step instructions

*         30-day free access to MARS, TreeNet, and Random Forests

More details:

*         Last week, we showed you how you could drastically improve prediction accuracy in your linear  regression with a new model that handles missing values, interactions, AND nonlinearities in your data.  This week, we will rebuild these original models and get straight to the more advanced features.

*         We will quickly review how to incorporate nonlinearities in a regression splines model  AND THEN show you how to automatically detect interactions and include these to lead to an even better result.

*         We will quickly review stochastic gradient boosting, and how, with plots you can see how each variable contributes to your model.   And then, this week you will see how to create approximations from these plots and use these in a standard linear regression as your inputs.

*         We will also explore the benefits of model automation. Without any custom programming, you can quickly cycle through different modeling scenarios, such as intelligently decreasing your predictor pool by removing variables one by one, or automatically re-running your regression model using different loss functions. This gives you the option to create many different models and choose the best for your analysis needs.



These techniques are great for skeptics who like to stick with standard regression but wish to see dramatic improvements. With very large datasets, you will see a significant speed benefit as well.  Learn what is being used at some of the largest banks and credit companies in the world.



And if you want a refresher, you can see last week's webinar at your convenience:


Who should attend:

*         Attend if you want to implement data science techniques even without a data science, programming, or even a statistical background.

*         Attend if you want to understand why data science techniques are so important for analysts.


Registration: http://hubs.ly/H01Y6dj0

Alternative Link: http://info.salford-systems.com/3-ways-to-improve-your-regression-part2

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.stat.unipg.it/pipermail/sis/attachments/20160126/c44c78aa/attachment-0001.html>