Generalizing Random Forests Principles to other methods

The full-length 42-minute video in HD 720p can be downloaded from http://www.crm.ugent.be/youtube1 (the slides are available for download from this same location).
This video lecture by Anita Prinzie (The University of Manchester, UK) and Dirk Van den Poel (Ghent University, Belgium) discusses our research about generalizing Random Forests (Leo Breiman, 2001) to other methods (both in classification and regression for data mining). We generalize our new method to Random Multinomial Logit and Random Naive Bayes. When applied to a case study in customer intelligence (cross-selling), we achieve a significant and substantial improvement in predictive performance over other multi-class classification methods in machine learning such as random forests and SVM (support vector machines).

Category:

Education

Tags:

• computer
• science
• data
• mining
• customer
• intelligence
• random
• forests
• machine
• learning
• multinomial
• logit
• (RMNL)
• naive
• bayes
• (RNB)
• AUC
• PCC
• Prediction
• error
• variance
• bias
• decision
• tree
• SVM
• Accuracy
• predictive
• analytics
• multi-class
• classification
• UGent
• Ghent
• University
• Belgium
• analytical
• CRM
• relationship
• management

[crchrdsn]

Random forests are fucking awesome. Too bad Breiman is not around today.

[antoniocherry]

Respect to Breiman, but bootstrap is becoming the miracle medicine to all the statistical problems. Too bad that it is so local.