README formatting

Update README. Code cleanup

Remove .pyc files that were accidentally committed

Added automated feature generation, removing highly correlated features, and plotting ROC curve

moved hyperparam optimization reporting to another file. Extensive tweaks to randomforest and started over in randomforest2.py

Implement PCA and KMeans clustering as data preparation step before training model. Acheived submission improvement to .794

Optimizing SVC->rbf model

Renamed linearsvm.py to svc.py to indicate that this will contain SVM implementations with kernels rather than linear, which is covered by the SGDClassifier code in sgdclassifier.py file

Loss function modified_huber is a tweak of hinge which has been best so far, so I tried that. Slight decrease in accuracy

Perceptron loss function didn't work at all, it couldn't even overfit successfully. Not submitting

SGDClassifier optimized as logistic regression. Updated to run multicore for faster hyperparameter optimization and learning curve generation at higher iterations

small updates to parameters for final tweak and submission

Updated SGDClassifier implementation and tweaked params for the 'hinge' loss function. Minor updates to naivebayes and random forests code

Lots and lots of changes. Refactor data prep into separate file. Update random forest to acheive new submission record .78+

experiments with sgdclassifier model, including a simple top-n voting ensemble. Ensemble hurt but single model improved my top score slightly