Fix broken inverse target transformation of LightGBM predictions

[HunterMcGushion]

• This change fixes experiments using LGBMRegressor with a feature_engineer that performs a quantile target transformation and inversion

• Steps to reproduce:

  1. Comment out the single line of code added by this commit
  2. Open hyperparameter_hunter/examples/lightgbm_examples/regression.py
  3. Add the following feature engineer step function:

  from sklearn.preprocessing import QuantileTransformer
  
  def quantile_transform(train_targets, non_train_targets):
     transformer = QuantileTransformer(output_distribution="normal", n_quantiles=100)
     train_targets[train_targets.columns] = transformer.fit_transform(train_targets.values)
     non_train_targets[train_targets.columns] = transformer.transform(non_train_targets.values)
     return train_targets, non_train_targets, transformer

  4. Add feature_engineer=FeatureEngineer([quantile_transform]) to the CVExperiment

• Following above steps should raise ValueError: Expected 2D array, got 1D array instead

• Seems that LightGBM is the only library to return 1-dimensional predictions given the opportunity; however, this problem could certainly be caused by some weird stuff I'm doing elsewhere...

[codecov[bot]]

Codecov Report

Merging #140 into master will increase coverage by 0.01%. The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #140      +/-   ##
==========================================
+ Coverage   91.06%   91.07%   +0.01%     
==========================================
  Files          43       43              
  Lines        4273     4282       +9     
==========================================
+ Hits         3891     3900       +9     
  Misses        382      382

Impacted Files	Coverage Δ
hyperparameter_hunter/space.py	98.01% <100%> (+0.08%)	:arrow_up:
...meter_hunter/data/data_chunks/prediction_chunks.py	100% <100%> (ø)	:arrow_up:
hyperparameter_hunter/optimization_core.py	92.51% <100%> (+0.1%)	:arrow_up:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 01ff723...4138eea. Read the comment docs.

[coveralls]

Pull Request Test Coverage Report for Build 608

• 12 of 12 (100.0%) changed or added relevant lines in 3 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.02%) to 91.079%

---

Totals
Change from base Build 604:	0.02%
Covered Lines:	3900
Relevant Lines:	4282

---

💛 - Coveralls