Directory and file contents:
- data/ folder contains only markdown file with brief decription of the data set as the original files are too big to upload them to github.
- mlcc_results/ folder contains results of MLCC executed on our data (clustering of variables, clusters' dimensionalities).
- notebook_results/ folder contains computation demanding notebooks executed using papermill.
- notebooks/ folder contains all notebooks used to generate reuslts from the thesis:
- Exploration and Normalizations include code used in section 3 of the thesis.
- DataPreparation includes code splitting our data to train and test sets.
- functions contains all imports and definitions of used models as well as all helper methods (plotting ROC curves, printing scores, wrappers for sklearn models, wrappers for models executed in R, etc.).
- Regularized and DimensionalityReduction include code for fine tuning models from section 2.2 and section 2.3 of the thesis.
- ModelSelection performs nested corss-validation for algorithm selection
- RandomLogisticRegression includes code for fine tuning and nested cross-validation of RLR model only. It was separated from other models due to long execution time.
- ThresholdAdjustment includes code for adjusting decision threshold and presentation of precision-recall tradeoff from section 4.2 of the thesis
- SAM performs Significance analysis of microarrays in R (section 4.3).
- ComputeSVCFeatureRanking contains code for Recursive feature elimination using SVC classifier (section 4.3).
- FeatureSelection includes code for comparison of feature selection methods from section 4.3 of the thesis.
- miscellaneous contains some helper code and a part not used in the thesis.
- selection_results/ contains results of SAM and RFE.
- config.yaml contains a configuration for running notebooks using papermill by run_notebooks.py script.
- environment.yaml contains exported conda environment in which the code was executed.
- mlcc_run_script.R is a script running MLCC on our train set.
The thesis itself can be found in the thesis repository.