Added EnhancedPredictiveModelBuilder interface, and EnhancedCrossValidator

the new additions make it possible to optimize data related hyperparameters, such as the decision to Log transform vs mean normalized numeric attributes. See the LogisticRegressionBuilderTests for concrete examples.

Made NDCG consider an instance's weight correctly

Merge pull request #145 from sanity/addNDCGWeights

Add ndcg weights

updated logistic regression data pipeline to use insufficient data labels for attributes that do not have enough data

Merge pull request #144 from sanity/addLogistic

updated date logistic regression data pipeline to use insufficient da…

Cleaned up logistic regression code

The code for logistic regression and data pre-processing has been cleaned up and re-organized.

Added Logistic Regression

Added an implementation of logistic regression that works with sparse attributes. Also provided data processing utilities for converting standard quickml instances to SparseClassifierInstances, which are required by the LogisticRegressionBuilder. See LogisticRegressionBuilderTest for examples.

updated the Classifiers class

Merge pull request #140 from sanity/updateEnum

updated version

fixed implementation of RankingPrediction to handle rank queries on items that are not contained in the ranked list

Merge pull request #138 from sanity/optimizer

Optimizer

updated RankingModel interface

Merge pull request #137 from sanity/optimizer

modified Ranking prediction interface

Added high level interfaces and loss functions / checkers for predictive models that do ranking.

Merge pull request #136 from sanity/optimizer

bumped version

updated default hyper-parameters of decision tree

0.9.5

bumped version