arpcode / smote-ada-boost-cobra

An Scikit-learn compatible mini-library with implementations of the COBRA (Classifier) aggregator (that takes any sklearn classifier as an estimator, as opposed to pyCobra), and also provides the functionality of AdaBOOST and SMOTEBOOST. Oversampled Boosting is introduced for dealing with Class Imbalance Issues

Cobra ROSboost for imbalance class problem

Dataset

pima-indian-diabetes:

file: pima-indians-diabetes.csv
--------------points
majority class 500
minority class 268

data:

file: data.csv
--------------points
majority class 950
minority class 50

Directory Structure

│   CoBra.pdf
│   README.md
|   SMOTEBoost results.pdf
│
├───Data
│       haberman.csv
│       pima-indians-diabetes.csv
|       data.csv 
│
├───Notebooks
│       Cobra.ipynb //implementation of COBRA
│       SmoteAdaBoostedCC.ipynb //implementation of AdaBoost, SMOTE and ROSBoost using SMOTE for oversampling
|       CreateDataset.ipynb //creates imbalanced dataset. data.csv containd data created using this code with 95% points in majority class. 
│
└───Scripts
        Boosting.py //implementation of AdaBoost, SMOTE and ROSBoost using SMOTE for oversampling
        CobraClassifier.py //implemenatation of COBRA
   

Results

SMOTEBoost results.pdf contains the results obtained. Results were obtained after applying ROSBoost using SMOTE on a cobra classifier made from 3 desision trees.




DISCLAIMER: This work is for learning purposes only. The work can not be used for publications or commercial products etc. without the mentor’s consent.