======= kdd2014

This is the best solution for team "Chaotic Experiments" led by Kiran R who is also the "sole team member"

#######Software Requirements###########

#######Data Mining Packages###########

Vowpal Wabbit (any version) must be installed and the command vw must be runnable from command line meaning vw must be in the PATH variable: https://github.com/JohnLangford/vowpal_wabbit
XGBoost must be installed from Tianqui Chen's repository: https://github.com/tqchen/xgboost
1. The development version of scikit Learn installed because we use GBMs with early stopping to prevent overfitting + we want to randomly choose the number of trees at each split
The following libraries must be installed in R data.table Matrix glmnet doMC foreach rbenchmark Metrics gbm RRF lme4

#####Instructions to Run the software####

Copy the files in github to your local machine
Set the variables in main.R correctly current_working_dir: Set to the current working directory where you place the above source files scikitLearnPath: Set to the installed location of the scikit-learn development version xgboost_path: Set to the XGBoost installed path
Copy the files of the competition and unzip them to the same folder as the source files i.e. to the current_working_directory above
The parts of speech features are painful to compute. So you may pick them up from here: https://www.dropbox.com/sh/fu9sx3jrdvtirlg/AAC_SrEeThn4-SxJBK2IsSzia
Run source ('main.R', print.eval=T, echo=T) to run the files

rkirana/kdd2014