Applied-Machine-Learning-Final-Project

General Information:

Team Name: Liad & Yuval

IDs:

ID #1: 300822954

ID #2: 311434047

This repository contains the following files:

1. Applied-Machine-Learning-Final-Project-300822954-31143407.ipynb - which is a notebook contains all relevant code. Allow to run the experiment and contains other sections of the project such as statistical hypothesis test, meta-learning-model, graphs, etc. Was tested on Colab.

2. flow.py - experiement flow controller + hyper-parameter search grid.

3. nested_cv.py - nested cross-validation infrastructure.

4. utils.py

5. rotboost.py - RotBoost implementation

6. rotation_forest.py - Rotation-Forest implementation

7. results/experiments_results.csv - our experiments results as depicted in the file 'Final-Project-Report-300822954-31143407.docx'

8. classification_datasets-20200531T065549Z-001.zip - holds all csv datasets zipped

9. /classification_datasets - holds all csv datasets

10. ClassificationAllMetaFeatures.csv - meta learning model's meta features

11. dataset-metadata.csv - holds datasets metadata such as binary/multiclass type, number of attributes etc.

12. results/feature_importance_weight.png - meta learning model feature importance type 'weight'

13. results/feature_importance_gain.png - meta learning model feature importance type 'gain'

14. results/feature_importance_cover.png - meta learning model feature importance type 'cover'

15. results/shap_summary_plot.png

16. results/shap_training_set_prediction.png

17. results/hyperparameters-search-space.png

18. Applied-Machine-Learning-Final-Project-300822954-31143407.docx

How to run:

1. Notebook - Applied-Machine-Learning-Final-Project-300822954-31143407.ipynb:

   Note: please use colab.

   1. Download notebook Applied-Machine-Learning-Final-Project-300822954-31143407.ipynb
   2. Download classification_datasets-20200531T065549Z-001.zip
   3. Download results/experiments_results.csv
   4. Download ClassificationAllMetaFeatures.csv
   5. Open Applied-Machine-Learning-Final-Project-300822954-31143407.ipynb notebook in colab
   6. The following files are needed to be uploaded to the notebook under '/content' dir (which is the default):
      • classification_datasets-20200531T065549Z-001.zip
      • results/experiments_results.csv (this file is not required if run_nested_cross_validation flag is set to 'True')
      • ClassificationAllMetaFeatures.csv
   7. Run all cells

   There are two options to run the notebook:

   • Skip the nested-cross-validation section and run only the processing results section + meta learning section. Default option.
   • Run the whole exercise (nested-cross-validation section + processing the results section + meta learning section). In order to that, please set in the settings cell (#2 cell) ‘run_nested_cross_validation = True’. Please note that this might take a while.

2. Directly: run flow.py file. Note: this option is less favorable and allow to run the experiment only, does not support other sections of the project, such as statistical hypothesis test, meta-learning- model, graphs, etc.