Testing Backtrack Search Optimization Algorithm as Ensemble Selection Method

Name of Quantlet : Ensemble_testing

Published in : Numerical Introductory Course (SS17)

Description : For testing of BSA ensemble selection method against stacking and individual classifiers.

Keywords : ensemble selection, ensemble search, search optimization, stacked generalization, stacking, random forest

Author: Shikhar Srivastava

Submitted:  Sun, July 23 2017 by Shikhar Srivastava

Datafiles: 
1. Two files from (https://github.com/shikhar-hu/Ensemble-Selection/tree/master/Datasets)
2. http://archive.ics.uci.edu/ml/machine-learning-databases/statlog/australian/australian.dat
3. http://archive.ics.uci.edu/ml/machine-learning-databases/statlog/german/german.data-numeric

Output : 'BSA performs better for 3 out 4 datasets with full library search. No improvement with pruned library.'

Procedure for reproducing the results (with using R 3.3.1 in windows 64it)

1. Create empty folder using below code

user=(Sys.info()[6])
Desktop=paste("C:/Users/",user,"/Desktop/",sep="")
setwd(Desktop)

dir.create(paste(Desktop,"/MEMS",sep=""))
dir.create(paste(Desktop,"/MEMS/S6/NIC",sep=""))
dir.create(paste(Desktop,"/MEMS/S6/NIC/Datasets",sep=""))

home=paste(Desktop,"MEMS/S6/NIC/Datasets/",sep="")
setwd(home)

2. Download this data-info CSV in the folder. Also download these CSVs into the same folder with same name.
3. Also download all the R-codes into the same folder.
4. Run the codes with ordered prefixes.
5. Take care that while running the codes under "3_" (which are classifier codes), you change the number of cores accordingly as it uses parallel computing.