R wrapper for sofia-ml suite.

If you want more details about the suite then ckecout the site or read the papers:

• D. Sculley. Combined Regression and Ranking. Proceedings of the 16th Annual SIGKDD Conference on Knowledge Discover and Data Mining, 2010.
• D. Sculley. Web-Scale K-Means Clustering. Proceedings of the 19th international conference on World Wide Web, 2010.
• D. Sculley. Large Scale Learning to Rank. NIPS Workshop on Advances in Ranking, 2009. Presents the indexed sampling methods used learning to rank, including the rank and roc loops.

Description

• code/doSVM.R R wrapper for sofia-ml suite. Before using the wrapper make sure you have installed the suit, and can run it from command line.
• code/validate.R computes accuracy, precision, recall, F-measure, predicted and expected distributions for a supervised model.
  • Supports case when prediction and groundtruth have different classes
  • Supports probability matrix (the class with the highest proabliry always wins) or probability vector (for binary models) or vector of predicted values
  • For binary models computes AUC-ROC and AUC-PR (using PRROC package), plots ROC curve, and finds an optimal threshold so that the predicted distribution as close as possible to the actual distribution
  • Can save results to CSV file
• code/generateSVMFiles.R given character array of data in SVM lite format and array of labels, function creates one file per class and returns character array of filenames.
• test/test_doSVM.R gives you a simple example how to use the wrapper. As a dataset I used IRIS dataset, see UCI Machine Learning Repository for details.
• test/test_validate.R tests code/validate.R function.

Citations

• Sofia-ml site
• PRROC: computing and visualizing precision-recall and receiver operating characteristic curves in R, for details see here. J. Keilwagen, I. Grosse, and J. Grau. Area under precision-recall curves for weighted and unweighted data, PLOS ONE (9) 3, 2014.