Statistical Learning

Basic description of this course

This course is part of a full series of statistics classes, organized by Genotoul - Biostatistics in the INRA of Toulouse during the years 2011 and 2012. Its aim is to introduce forecasting methods coming from machine learning. It is divided into

• a presentation of the topic (approximately 3 hours);
• practical applications using the free statistical software environment R .

The course is organized as follows:

• basic introduction to machine learning;
• introduction to multilayer perceptrons (neural networks);
• introduction to classification and regression trees (CART algorithm);
• introduction to random forests.

The following R packages are used in the applications:

• car (usefull tools);
• nnet (neural networks);
• e1071 (machine learning);
• rpart (classification and regression trees);
• randomForest (random forests).

2012 material

• Slides: theoretical part and practical application;
• data: a list of selected genes expression, a list of phenotype values (pH), a split between training/test data sets, a list of selected genes;
• R scripts: script file.

How to use the material?

Once all the files have been downloaded, the best way to use them is to make the following directory hierarchy:

• a folder named “data” containing the data (Rdata and csv files above);
• a folder named “ML” containing two subdirectories:
• one named “Cours” with the theoretical part slides;
• one named “TP” with the slides and the R script related to the practical application.