dcndn4 / HW_8_ML_Conf_Imb_U11

Unit 11 Homework - Risky Business - ML w Imbalanced Lib and Confusion Matrix

HW_8_ML_Conf_Imb_U11

Unit 11 Homework - Risky Business - ML w Imbalanced Lib and Confusion Matrix

Resampling model process

I examined the LendingClub data with multiple algorithms, to determine which worked best for this data set.

The these included a simple logistic regression, which yielded the following results: balanced accuracy score: 98% Confusion Matrix: 615, 10; 102, 18657

Also an oversampling process: naive random oversampler Confusion Matrix: 622, 3; 111, 18648

Also oversampled with Smote Confusion Matrix: 622, 3; 110, 18649

Then I did an undersampling process with ClusterCentroids, and the result was: balanced accuracy score of 99.4% confusion matrix: 620,5; 102,18657

Finally I did a combination (over and under) sampling process using the smoteen algorithm: balanced accuracy score: 99.32% confusion matrix: 622, 3; 110, 18649

The best accuracy score was the undersampled one.

Ensemble Learning

I applied the Balanced Random Forest Classifier and the Easy Ensemble Classifier to predict loan risk and evaluate each model.

Technical Notes

Libraries

This Jupyter Lab notebook utilizes the following libraries:

-- path

-- collections

-- sklearn

-- imblearn

-- pandas

-- Numpy

-- Pathlib

-- matplotlib

Acknowledgements

I would like to first acknowledge the guidance and teaching of our FinTech Boot Camp Instructor, Garth Mortensen, our TA, Alejandro Esquivel, and out Student Success Manager, Angelica Baraona. I also found the collective Stack Overflow wisdom (especially the Quantitative Finance community) essential as ever. The confusion matrix plot was provided via the website: Notes on Artificial Intelligence, Machine Learning & Deep Learning, Python, Mobile & Development, by Tarek Atwan. Regarding Machine Learning overall I utilized the book "Hands-On Machine Learning with Scikit-Learn, Keras & TensorFlow" by Aurelien Geron.